Cynulliad Cenedlaethol Cymru

Good morning, and welcome to this meeting of the Culture, Communications, Welsh Language, Sport and International Relations Committee. We have received apologies from Lee Waters and also apologies from Heledd Fychan—Sioned Williams is substituting on her behalf. Are there any declarations of interest from Members, please? I don't see that there are, so we'll move straight on.

We are looking into our inquiry on 'Cymraeg for all?', and we're very pleased to have the panel with us today. I'll ask our witnesses to introduce themselves for the record, and I'll go to Lowri first of all.

Good morning. Lowri Hughes, head of policy and development, Canolfan Bedwyr, Bangor University.

Thank you. And Gruffudd.

Good morning. Gruffudd Prys, head of language technologies, Canolfan Bedwyr, Bangor University.

Thank you very much. And Llion.

Good morning, Llion Jones, director of Canolfan Bedwyr at Bangor University. 

Welcome to all three of you. If it's okay with you, we'll go straight into questions, Every one of you doesn't need to answer every question, but, whichever one of you wants to come in, if you want to put your hand up.

In terms of the current tech and digital landscape that we have at the moment, where do you think we're at in terms of the Welsh language? What does that landscape look like? 

I think that we're in a really good position in terms of the size of our population, the number of speakers that we have, and the fact that Welsh is part of this AI revolution from the outset. If we look back at the past, when even 'ŵ' and 'ŷ' didn't exist in our software, it's good to now be included in these large language models from the outset.

I think that we're a little behind the curve as compared to English, but that's true of a great number of other languages too. I think that what we need to ensure is that the data for the Welsh language are available to feed in to ensure that the ability of the Welsh language side of these large language models is as good as that of the models in English.

Thank you for that. Llion.

Your use of the word 'landscape' is quite interesting there. When I think of a landscape, I don't often think about people. But I think what's important in terms of technological developments in the Welsh language is that they put people at the centre of them and that we put an emphasis on those technologies that will enable use of the Welsh language. It's important, I think, that, when we develop things from scratch, we do do that with an eye on the context in which people use the language.

That's interesting. Thank you. We'll move to Sioned.

Thank you very much. The Welsh Government's 'Welsh language technology action plan', the objective was to ensure that the Welsh language was central to the experience of technology. To what extent do you believe that that ambition has been achieved in terms of innovation in the Welsh language?  

I was looking at that plan the other day and I could see how many of the objectives in the plan had been completed or realised at this point, even though it is quite an ambitious plan, especially in terms of the infrastructure element of building and preparing those foundations for the technology. At the time, no-one had foreseen how fast AI would develop, but I think what had been realised at the time was this pattern of needing a lot of data and a lot of building blocks—that was something that had been realised. So, that work had been started. 

I have to acknowledge that we in the language technologies unit do receive funding from the Welsh Government as part of the package, and we have been responsible for creating a number of those resources. But we do feel that that aim has been reached in terms of the building blocks themselves. Where a lot of work is needed is ensuring that those reach the users and their daily experience of using technology. A lot of that is happening within software packages that have been prepared and commercialised by major international companies, and we need to ensure that those companies, and the people who buy them as well, truly respect the language standards that we have here in Wales already, and that they understand the consequences and what's expected of them in terms of linguistic support in those packages, because the existence of those building blocks means that there's no excuse to not have Welsh text-to-speech and speech recognition features within these packages at this point.

You mentioned that you'd received funding from the Welsh Government. Would you consider that that funding has been sufficient? And in looking ahead at the new priorities of the Government in terms of Welsh-language technology, do you believe that the funding and the priorities are set at the right level? And could you expand perhaps on what the role of Canolfan Bedwyr is in that regard?

So, at the moment, we receive around £350,000 a year for a specific annual project. That is still based on the world as it was back at the start of the plan. I think what has happened now is that we've realised how much more crucial language technologies are than was considered at the time. So, I would argue that we need to increase that grant significantly at this point, considering how vital these technologies are to everyone's experience across a wide variety of fields.

One of the other things that's a bit of an issue for us is the annual nature of the funding. I think that more long-term and better planning would be possible if we moved to a three-year model of funding this in the future.

As the official in charge of managing the funding, I would second that, because we are very fortunate. We have very talented and able developers who are committed to the Welsh language. They could have very prosperous careers in other places. So, the ability to sustain that workforce, to maintain it—. But also I think we need to look to the future and develop developers who are immersed in this as soon as possible.

In terms of that, do you have any problems in attracting people? Obviously, there's big money in this field. As you mentioned, people can earn a lot, and it's good that we've got people who are committed to the Welsh language. Do you have any problems in recruiting or are you thinking about the sustainability of the workforce for the future in terms of this crucial work?

It is challenging. One of the things that we do in addition is provide a Master's degree in language technologies to ensure that there is that new generation coming through. But we also have to ensure that our current workforce is content and that there is a future for them in this industry, and short-term contracts to militate against that in terms of providing job security for people who are at the heart of this technology for the future.

I think, in terms of the behaviour change project that's been developed as an app now, we have also developed expertise in the behavioural change and language planning field. That also needs to be sustained and developed over the long term. You were asking about Canolfan Bedwyr's role. We can, of course, bring together all those language planning elements: the technology, the policy, the translation, tutoring and so on. And that, in a way, is responsible for the innovation very often—bringing all of this stuff together leads to innovation. But we need to sustain that and maintain the expertise, and provide an opportunity for all of those disciplines to be developed within a holistic context, where it's possible to trial things with a variety of linguistic contexts, as we have in a university, as it happens. That then can be attributed to various linguistic contexts across Wales.

Going back to that action plan, we're looking back over the years of technological developments, and there are several different projects and different action plans that have been developed but haven't been maintained and developed. And so I think that having the robustness of a number of people working across a number of areas helps in terms of the sustainability of those projects.

But I would agree with Gruff that what the action plan has delivered is exceptionally good, bearing in mind how changeable technology is, and bearing in mind that lockdown landed in the middle of that action plan. Providers did adapt very well to that, I think, and prioritised things such as subtitles and so on that were required as lectures and so on became things that happened online

Therefore, could I just ask you generally whether you believe that the new priorities of the Welsh Government for Welsh language technology do provide the correct foundations for the future?

In general terms, yes, they appear to be sensible, but the world is moving so fast in this area at the moment that it's difficult to forecast where things go next.

So, can I take it from your answer that flexibility is needed, and that this plan needs to be a flexible one that can evolve? 

Yes, and I think that the emphasis on data and open model packages is helpful as well, because that means that the work is not tied down to a particular ownership and that the data and the research work that's gone into that is in everyone's hands—all stakeholders—and continues outside of one provider. Perhaps the concern that we might have in the future is that we are perhaps too dependent on one commercial international provider. Therefore, we need to ensure that we fund that the infrastructure below that, independently of any specific commercial provider, so that we can move from one provider to another, and develop competition in the area as well, so that there is no temptation to raise prices on us down the line.

If I could go back to your question around Canolfan Bedwyr itself, I referred in a previous session to the way that we try to work within the framework of the Government's Welsh language strategy. That is, there are three criteria—creating new Welsh speakers, encouraging use, and infrastructure—and I hope that we have a grip on all of those, and I think that working in an integrated way is key to the success of the strategy.

Thank you. Thank you, Chair.

Thank you. We'll move to Alun.

Thank you. I'm enjoying this discussion. It is very interesting. I assume that—. As we move forward, you say that it's difficult to know where this technology will lead us. I agree with you on that. There are some things that I've seen that I never expected to see. But there are things that we do need to understand as well, and the aim or the objective is that the Welsh language will be a central, essential part of any kind of new technology. So, we will have the right to speak Welsh with Alexa, and, with the way that these interfaces are going, I use Welsh on my iPhone, and I have started using Welsh with Copilot. So, how do you see artificial intelligence changing how we work and also communicate? What should the Government be doing in order to ensure that we reach the objective that you've described, that the Welsh language is a vital part of whatever future happens?

Well, one of the areas is creating those standardised evaluations, setting out what we expect in terms of the quality of the AI. When you use Copilot or OpenAI or Anthropic, the Welsh language is surprisingly good. It isn't perfect, and it doesn't reach the standards that we would expect from an author—

Well, my Welsh isn't perfect either.

That's true to say, but there's a difference of use, isn't there, between professional use within our organisations and facilitating Welsh speakers to use more of their Welsh language skills. What I would say is that we need to—. We see that with speech recognition, specifically, that we now have to deal with people's everyday Welsh, warts and all, and the fact that people do throw in these English sentences into their speech. So, the evaluations have to be able to reflect that.

I'm not saying that the evaluations necessarily are about writing something that doesn't include any errors according to Cysill. These evaluations include being able to deal with English words spoken as part of Welsh, but also deal with bilingual names. For example, Rhos-on-Sea was translated into Rhos ar y Môr for a few months, and there are cultural considerations that are specific to Wales as well. We need to ensure that the AI can respond to those.

So, speaking about the evaluations, they're linguistic, cultural, sociological evaluations, ensuring that they don't produce hate speech, for example, that kind of thing, but there's also a question about how one would do that in Welsh in real terms, not using those formal, verbal forms, but perhaps using Welsh forms for English swear words, so dealing with how we actually, really use the Welsh language from day to day. 

I know that some people perceive this technology as a challenge. I see it differently, as a bridge, and I think it's important that we're part of it. I represent Blaenau Gwent and I don't speak a lot of Welsh in Blaenau Gwent. My son lives in Hay-on-Wye and he doesn't hear any Welsh apart from when he speaks to me, and like most teenagers, he doesn't want to do that very frequently. [Laughter.] So, how can we use this technology as a sort of bridge to be able to enable people, where they don't hear a lot of Welsh from day to day, to actually use the Welsh language and expand the Welsh language in an area where there isn't much natural Welsh being spoken, if you like?

I think that the technology offers an opportunity for us to introduce the Welsh language and to demonstrate that it's something that is possible for people. That's what's behind the ARFer change in language use project that we have as part of our work. I think it's an opportunity for them to hear the language, to see the language where they perhaps wouldn't otherwise, but there's a next step. The fact that something exists doesn't mean that young people, for example, are going to use it. So, we have to look around that.

Let's take a workplace, perhaps in Blaenau Gwent, for example. It's a way to introduce the Welsh language to the workplace for some speakers—'Well, actually, I can understand some of that—I can speak a little bit of the language.' But then we need to look at the leadership structure within the workplace, how much credibility is attached—that is, how much input is placed on the Welsh language in terms of structures, in terms of procurement processes and so on, how much value is attributed to the Welsh language. So, there's an element of understanding those wider structures around the users of the Welsh language, I believe, and I think it's going to be a risk in developing technology without looking at the surrounding support structures, and the context of people.

Copilot as well, or similar technology, can be a friend that can speak Welsh with you. You can practice your Welsh with Copilot without judgment, without having to do that publicly, so that it can help you reach a point where you might be more content to use your Welsh language in a more public setting afterwards. That is certainly something that we need to look at.

Yes, I see that. Also, when I was elected almost 20 years ago, when we were discussing the Welsh language, we were talking about the translation of books, basically. I don't remember the last time—. I don't think this committee has discussed that once in the past four years because the research is now all done on the web. We don't go to libraries like we used to. What is on the web is mostly in English, and that's true for almost every nation. So, how can we ensure that there is an environment, or a bridge, in the Welsh language where everything that we see around us is in the English language? How can AI help with that Welsh environment, do you think?

One of the things that's quite amazing in terms of AI—. I was reading a paper by Anthropic recently and they were suggesting that the models had started thinking beyond language, above language in a way, or beyond it, and that they had then started generating the responses in the language that was asked for. And what I saw when using AI to ask for information for another presentation about Bangor pier was that Claude by Anthropic had produced a better explanation of the history of Bangor pier than what was in the Wikipedia article on the Welsh Wicipedia website, and Claude had produced it in Welsh. So, there is a future where we might not do our searches online for information, but we will go straight to AI to get it, and that would then be transferred from the concept to Welsh text or to a Welsh voice.

So, the gaps I perceive then are where we possibly might not have the terms in Welsh for the kinds of fields—. We could be asking a question in Welsh and getting a response from AI in Welsh about it. So, I've been talking about examples like the pieces or the components of a mountain bike, and we're not all agreed on what some of those technical terms are for some of those bike components. And what Claude or Copilot would do then, perhaps, is that sometimes they might use an English word and sometimes would try to invent a Welsh word that has never existed previously from individual elements—sometimes it works, sometimes it doesn't work. So, we have to ensure that that vocabulary is there.

So, they're like T.H. Parry-Williams, just coming up with new Welsh words all the time. [Laughter.]

We've seen examples of that. I saw 'manetiau' being suggested as a word for 'handlebars' at one point. It looks like a Welsh word at first glance, but yet it's not really a Welsh word. But there is a danger of that happening as well, and the way to get around that is to create more data online in Welsh on Wicipedia, to have magazines discussing motorbikes in Welsh and things on television. There are Tour de France programmes on tv, but as far as I know, we don't have a lot of programmes about mountain bikes, for example. That vocabulary is going to come from real discussions and resources that are commissioned, often by the public purse.

And the question of young people is very important in terms of the platforms that those young people use. This technology needs to find its way to those platforms.

Yes, and I'm thinking now about games on the Xbox and so on—the kinds of things that I frankly don't understand. How do we ensure that the interfaces, if you will, are available in Welsh, because somebody else playing in another place can have an interface through another language? I take it that it is possible to do that. 

One of the things that I wanted to ask you about, and one of the things that the committee has looked at, is the use of services available through the medium of Welsh. Over the years we're all familiar with the campaigns that have led to greater use of the Welsh language with the Government, for example. It's possible now to fill in tax forms in Welsh, but we all know too that very few people actually do that, because of the complexity. It's very complex regardless of the language that you use, and every one of us will have something in our brain that compels us to use English when we fill in those forms.

So, how can you use something like AI to help people, to give them confidence, to use more of their Welsh language skills when using services? Because I take it that more services are going to be available online next year as compared to last year, and even more will be available in five years' time. So, the interface is going to be more of an online one than a paper one. So, how can we use AI to offer people more confidence to use their Welsh language skills in that context?

I think one risk that there has been for the Welsh language over the years is that it has been seen as something for just some specific contexts. I think that AI and technology provide an opportunity to open up the Welsh world out to everywhere. You were talking about interfaces on computer games, but perhaps the most important thing is what language they use when they chat together online. I think that the informality of the online world and AI is perhaps a way of opening the doors and raising confidence, and seeing people, as you were saying, Gruff, just trying things out in less high-risk situations. So taxes online—perhaps that's somewhere people perceive a risk, and it's something very official. So, perhaps that would be a step too far, but there's a more fluid technological world that we're living in at the moment. 

But we want people to use those services that we have fought to get. 

Of course, but perhaps there are steps to take them there, and that it's a journey for people. Perhaps AI—I don't know enough about the detail—will create a more informal context for people to try things out, and that the language—as you mentioned code switching, the language that reflects real language—will become easier to create in this virtual context. But I would say again that we are only going to lead people to it deliberately. We have to be deliberate in terms of what we do, and not just expect people to adopt things without there being guidance, whether that's on a national level, on a community level or even within a family. We have to think about the structures around things.

Thank you for that. David Crystal has written in English that there are many different Englishes. Because we are looking at the Welsh language and what needs to be done to promote the use of the language in those areas that Alun has spoken about, such as in the Valleys, where perhaps the density of Welsh speakers is lower—. In terms of the different Welshes that exist in these places, is there an additional challenge, if we look at something like AI, to ensure that everything doesn't become one kind of Welsh?

That is certainly a challenge. This technology is based on training data. A lot of the Welsh training data that exists on the internet is quite formal Welsh, official Welsh. That's where I think that the voice recognition technology that can, perhaps, turn radio broadcasts and podcasts that are less formal, and engage more with daily experiences, rather than dry official documents, are vital for us to be able to collect and convey the sound of people speaking in Welsh into Welsh text that can be evaluated and used more broadly, and once again to train a further generation of AI language technologies.

I do feel very recently that the speech recognition technology in Welsh has crossed a bridge, in that it is of excellent standard. But we need to spend more time improving that to ensure that all of the dialects in Wales can be identified correctly, and that we don't over-standardise, but that we have some sort of consistency in terms of how we spell words like 'moyn' and 'actually' as well—these English words that we use in Welsh. How would you want to see a word like 'actually' written down in Welsh? 

Or 'cwtch'.

Or 'cwtch', yes. Do we want the 'ch' at the start of 'checio', or can we learn to accept 'tsi-', which isn't a letter in the Welsh alphabet, but perhaps by seeing it on TikTok in the subtitles people would come to recognise that? So, we need to have a discussion like that about perhaps not the Welsh of today, but what we want to see in terms of the Welsh of tomorrow that would reflect all of the dialects and kinds of Welsh, and that linguistic mix that there is here too. 

Interesting. Sioned, you wanted to come in. 

May I just come in on that? As somebody who comes from the Rhymney valley and whose family spoke the Gwenhwyseg dialect, I had a Welsh-medium education and it was—it wasn't beaten out of me at school, not literally anyway, but I wasn't encouraged to use the dialect that I had naturally. We don't want that to be repeated, do we, with this technology? It would appear to me—I don't know whether you'd agree—that this is an opportunity in terms of using the speech recognition technology to safeguard these different dialects for the future. They have been under threat, haven't they, through the education system and perhaps through the way that we translate into Welsh, for example, and present official documents. Do you think that that's an opportunity? 

Certainly. It's exciting, in a way. It has safeguarded an element of the Welsh language already—we can acknowledge that—and it's important that we then ensure that the Welsh language doesn't become some sort of an exhibit in a museum, where you have AI that can speak Welsh, and that we have stopped speaking Welsh. That is something that we need to avoid, isn't it, obviously. But I think what we've tried to do with the approaches that we've taken is we transcribe them verbatim, what we hear, and we try and map that then to standardised forms, so that we can support 'jolch' and 'diolch'. We could change 'jolch', with a 'j' at the start, into a more standard form, if there is a demand for that, perhaps in order to translate using machine translation into English. But there are sociological questions relating to what we choose in terms of the form of Welsh that need to be answered.

Yes. I know that we'll need to make progress, and thanks for your patience, Mick and Gareth, online. We will move on, but I just want to ask one further question, because Alun and you, Gruffudd, have talked about this idea that AI can be a bridge, but sometimes there are trolls underneath the bridge, in terms of some of the challenges. I don't know what the word 'ethical' is in Welsh—

'Moesegol'.

'Moesegol' in Welsh then. How can you overcome some of those challenges, to ensure that the opportunities are ones that we take advantage of and that we don't fall?

We have to use AI appropriately, don't we, and ensure that that's the way forward. So, we are working on creating evaluations that can identify hate speech, as I've mentioned previously. That is incredibly important. But there are sociological questions there too in terms of how people are insulting in a realistic way, because it's complicated, isn't it, and we don't have a lot of that in our traditional dictionaries, perhaps. So, it's something that we're scratching our heads about, but we are addressing that question.

I think, yes, in terms of synthetic voices and to what extent we're willing to accept Welsh language services that are artificial, in a way, I think that it's important in cases like those to ensure equity with the English language services. So, if there's a use of English synthetic voices, then it would be okay for us to use synthetic Welsh voices, but I wouldn't like to see a future where Welsh synthetic voices are used where that isn't the case in English.

I think that there is a genuine risk there that we could be exempting organisations from creating workforces that can use the Welsh language naturally too. So, that's a risk.

Certainly, there's a place for investing in people and their skills as well, hand in hand with the technological developments, because that, in a way, is what's going to keep it real and alive and relevant and useful for us all, in a real way, perhaps.

And there is a need for more refinement in terms of the language standards then to say what's appropriate in what context, in terms of the use of technology, isn't there?

Thank you for that. We have to move on, but I'm sure perhaps there'll be some other things we will want to discuss with you after the session, if that's okay with you. We'll go to Gareth next.

I can respond to the first part of that question, I think, by referring to the app that has been hugely popular by the language technology team. The Ap Geiriaduron, the dictionaries app, perhaps doesn't sound very exciting, but it's been downloaded 400,000 times. When you consider that in the context of the population of Welsh speakers, that's a very significant number. So, I think being involved in something like that that's accessible, that is easy to use, and that is vital—people need to know what is the Welsh word for this, that, and the other. So, I think that it's very much related to the use of the app and how useful the app is. I wouldn't consider the dictionaries app to be an educational app in and of itself. Obviously there's a wealth of information behind it, but it is very useful, and I think that that is the challenge: finding those apps that will facilitate people's lives and their work through the medium of Welsh.

Yes. And there's power in the educational sphere to be distributing apps like the dictionaries app. We've seen quite a lot of those installs coming through educational institutions, and we need to capitalise more on the educational sphere as a means of distributing Welsh-medium apps. However, I also feel that those apps have to be easy to use and interesting, and that that is an important part of promoting Welsh-medium content through apps.

I think it's also fair to say that there are apps like Duolingo that aren't necessarily Welsh language apps, but they have encouraged the use of the Welsh language to an extent, or have increased interest in the Welsh language. As far as I know solely Welsh language apps are relatively rare, but technology provides a way of introducing Welsh within apps, for example, the days of the week in some of your shopping apps and so on. So, it's certainly possible to introduce the Welsh language through apps, but I agree that when we invest in them, they need to develop the use of the Welsh language and support that use too.

Gareth.

Thank you for that response.

Yes, certainly there are fewer excuses now for not providing services through the medium of Welsh. I was speaking with the head of a translation company recently who explained to me how easy it is now to translate the software interface, and that's what we're talking about here in truth, as a banking app wouldn't be very Welsh, they would change the—. We're talking about just changing the labels that appear on the screen into Welsh ones, and that is quite a simple process within the translation software packages that are available to translation companies. At the end of the day, it all comes down to the standards and whether specific organisations fall under the language standards or not. If only commercial considerations are taken into account, then often we see that it's not a priority for international commercial companies. 

Thank you for that. Okay. Thank you for that, Gareth. We'll go to Mick.

Thank you. A very good question. We have many, many colleagues in the other Celtic languages. For example, we speak often to colleagues in the Basque Country and Catalunya. There are things that we have in common; there are many considerations that are also different. The Welsh language is quite different to the Basque language, for example, so there isn't as much in common there, but perhaps in terms of the way that we are trying to restore our linguistic communities, we have a great deal in common with the Basque Country in that regard.

I think that we are in the vanguard when it comes to many of the other Celtic languages, and we are in the cohort, like the Basque Country and Catalunya, in terms of the technology that is available. We work closely with Ireland, but we also provide dictionaries in the Cornish language and we provide speech recognition for Scottish Gaelic, in collaboration with the University of Edinburgh. So, we learn lessons from each other. We learn from each other's mistakes as well at times, but often the political sphere and the nature of our languages are quite different as well. And sometimes I feel that there is a danger for us to be overly romantic and to turn to our sister languages, whereas really we are learning a great deal from the Anglo-American mainstream use, rather than pursuing those things that sound more like romantic connections. Those are very, very important to us, but the technological developments have been happening through the medium of English, and we have to acknowledge that. 

Perhaps one area we can look at—there was a question about apps earlier—is the data that comes from this use of technology, so that we can start to understand language use patterns and when the technology is being used, where, and in what linguistic context. And perhaps there is potential there in terms of models of use that we could learn from, and also for us to teach others internationally about that and to build on that.

May I just also say that we use systems from Ireland, from Dublin City University, to develop our terms? We've also collaborated through the Agile Cymru project with the Bretons to develop live speech recognition. So, in real time. So, there are very important connections there. 

Thank you. I think that Llion wanted to come in. 

Just on Lowri's point, as it happens, the new app that we have out now, ARFer, the methodology behind that one comes from an intervention in the Basque Country—this idea of having a leader within a group that facilitates the use of the Welsh language. But in that case, we have taken that idea and taken it in a slightly different direction. There are lessons to be learned from these partnerships.

And I think that the live data, Gruff, is important, because at the moment we tend to undertake linguistic assessments by looking back at language use, or perhaps through people forecasting language use, but how people are using the language at the moment is very important for us to collate and understand.

Mick.

Well, I think—I referred to Duolingo. There are a lot of people, of course, who learn through Duolingo but can't actually speak the language, and that is exactly what the ARFer app that we've developed tackles, in that it helps that language use element.

We have a team of language tutors in Canolfan Bedwyr in the university who have developed a system of language levels, and that has been developed on the basis of their experience of how difficult people find it just to start sentences, to take what they've learned in the classroom and to turn that into speaking the language with someone else.

We have taken a lot of that work and we've incorporated it within the tasks in ARFer, because I think we need to acknowledge how difficult it is to just break that barrier of hearing yourself using the language, and, in the same way, that speakers who might be more fluent feel comfortable supporting and speaking in a way that people who are less confident can understand them.

That's a process, I think, that does need care, and we need to think carefully about that. It's not just technology for technology's sake. That technology needs to be developed properly, you need to have that understanding of the language context that's behind it and an understanding of how people act together linguistically behind the scenes so that it becomes a truly valuable technology to change our language use patterns.

I have mentioned live data, I think that's important. I think understanding the interactions between different life experiences is important. For example, if we take children, what's the influence on them at school, what's the influence in the household? What's the interaction between that in terms of their clubs and so on? The same applies in the workplace.

There is a great deal of mention made of defaults in the field of behavioural change, and that's what we see at the moment in terms of the Welsh Language Commissioner's report. Young people default to the English language, particularly in the context of technology. How do we change that practice? I think we need deep research into how we change that default.

That's related to the technology itself, but it's also about children and young people's understanding, empowering them, providing opportunities to them through technology to change how they see things, and the defaults as well. We talk about nudges and so on, but, as I said, any intervention that is developed needs to be based on understanding of the linguistic context, the linguistic requirements and how people use language.

Just on the back of that, I would add that developing technology in a real context is vitally important. I think that's where we've been fortunate, because we've been developing our technologies in a university where we have this whole range. We have teams where not much Welsh is spoken, we have teams who are completely fluent, we have confident Welsh speakers but we also have more reluctant speakers, and there's the 'can speak, won't speak' phenomenon. So being able to trial things in this managed context is a way of coming up with solutions that address some of these issues.

I was with Mick in Kyiv a few months ago. We had a meeting with the Ukrainian language commissioner, and the conversation that we had there was very similar to the conversation that we're having now and that we've had in Wales. The conversation was about translating from Russian and English to the Ukrainian language and then the use of the Ukrainian language in technology and everything else. It struck me that there are many countries having the same kind of conversation outside of the English-speaking world.

You've touched on how we work with some countries or different language communities. I just wanted to ask you how we can collaborate more closely with other language communities, because we're talking about the 0.5 million of us who speak Welsh, and fewer people speaking other languages, but together, there is a large group of people who want to use a language that isn't English, German, Russian, Spanish or French—these major languages.

Together, I would take it that we could collaborate, because computers and programmes don't speak any language, they just do what we ask them to do. So, I take it that there is capacity to do more together than separately. 

The best example of that would be the Network to Promote Linguistic Diversity. I think that members from Wales are on that and they do meet every three months and share experiences, share good practice.

But sharing programmes of work—yes? Because I take it that Amazon or Google don't mind whether we use Welsh or a different language, as long as we can work together with the kinds of technologies that they have.

I think what's common between us and Ukraine, perhaps, is that there are two languages that are used side by side. I think perhaps what's a problem for us frequently is having to change the language of a spell checker in Outlook or to switch back and forth. You install a new piece of software, it's available in Welsh, but you have to go in and you have to change the settings to Welsh every time.

There's Welsh on Android phones, for example, but Samsung hides where you can change that. You have to go into 'other languages' to see that that's there. There is room for us to work together, to collaborate with those countries where we have more than one language to say to these international companies that it needs to be easier to support those two languages at the same time within software.

Fortunately, that's one of the areas where the AI seems to excel—you can change back and forth between different languages with Copilot, for example, and it does it straight away. So, perhaps we're moving in that direction already. I think that we need to form coalitions as bilingual countries together to insist that that happens as well.

This has been so interesting—thank you so much to all of you. In the last few minutes remaining to us, is there anything that you'd hoped to discuss this morning that hasn't arisen? We've raised a number of areas, I know. 

One of the barriers that Gruff's team in particular faces, I think, are some of these major companies, it's so difficult to incorporate our technology in those now. We referred this morning to the issues that have been with Cysill software as Microsoft tightened up the ability of external developers to develop add-ons within that.

Yes, certainly, and I think we need to congratulate Government officials who have ensured that Microsoft has included much more of the Welsh language in its software over the past few years, but there are other companies such as Apple, for example, that don't seem to be responding at all at the moment to the need to be including the Welsh language. So, I feel that we need to be using our purchasing power to place additional pressure on companies such as Apple and say we won't buy thousands of iPads for Welsh schools unless they can support the Welsh language better.

The other consideration that does cause a problem for us is that we are trying to produce open building blocks under open licences, we are asking the public sector to share their data with us so that we can package those data for ease of use by a range of different organisations. We often run into issues with data protection and GDPR in all of this. These are issues that could be overcome, but, unfortunately, public organisations have data protection officers, they don't have data sharing officers.

I would like to see a change of mindset within the public sector, where they can balance the need to be sharing data with us and with these major international players to improve the models, whilst also safeguarding confidentiality and personal privacy and so on. I think that we can strike that balance and do that. But at the moment, the penalty for releasing private data is so significantly greater than breaching some language standard. And that means that people are very reluctant to share as much as they could. 

So, we need to put processes in place for the future that facilitate that from now on. Because perhaps with the data in the past, we can't go back in time and get everybody's consent amongst those who contributed to those data, but if we could organise the way that we tag public data in future and say, 'This is appropriate to be shared, this isn't', then perhaps we can provide more Welsh data and make that available to be used to improve the standards of these models.

Excellent. You've given us a lot to consider there. So, thank you very much for your evidence. A transcript of everything that's been said will be sent to you for you to check that it's an accurate record. Perhaps there will be a few things that we want to follow up with you on in writing, if that's okay. There might be some additional questions, but thank you very much. That was incredibly interesting this morning. Thank you very much to you for your time.

Members, we will now take a short break. We'll start again at 11:50 with the next session. We'll go into private session now. 

Good morning and welcome back. We are moving on to our second evidence session this morning. We are joined by Carl Morris to give us evidence, and I will ask Carl to introduce himself for the record.

Excellent. I'm Carl Morris. I work as a Welsh language technologist. I have around five minutes of themes to share with you to begin, if that's okay. 

Yes, that's fine.

This is just to give you a bit of background, for those who might not know me. So, good morning and thank you for the welcome. I am working on a variety of projects in the technology sphere, in relation to Welsh, with a variety of institutions, on a freelance basis. So, I don't represent any particular group or institution, apart from myself. The first session this morning was very interesting. Technology is an enormous field. I concentrate primarily on ICT, including the internet, apps, phones, software, computers and digital media.

I'd just say at the start that there are some areas of technology that are relevant to the Welsh language particularly, like agricultural technologies, for example. Don't ask me about those; I'm not an expert. But, just to note, technology can mean the use of information to achieve a task or to solve a problem. Almost anything that people create is some kind of technology. Shoes are technology, for example. But I'll try and concentrate on my field of expertise in ICT today.

I had an opportunity recently to work with Mudiad Meithrin, to look at opportunities to use technology to support Mudiad Meithrin as a body, but, more importantly than that, the cylchoedd meithrin groups. The question in my mind was how can we use new technologies to support what's happening in the cylchoedd meithrin already—the immersion technique or approach of creating a Welsh space where children under five, in some cases, get their first taste of the Welsh language?

I would argue that people come first and the things that we want to do come first, and that technology then serves that aim. So, we have identified that toys are a very important field. Mudiad Meithrin has listed Welsh toys of all kinds that are on the market at the moment that are appropriate for children under five years of age. Those include toys that are not electronic, some that are electronic, and then some that sing and speak in Welsh, for example. I would be happy to expand on that later on.

As it happens, the Welsh Language Commissioner has published a brand-new report on children's practices in terms of how they use Welsh. Perhaps this is more appropriate as we move towards school age—primary and secondary, and so on—and I'm happy to expand on that. It sets a challenge for anyone who is concerned about the Welsh language in the technological sphere, and in the digital media and social media. The responses from young people are very interesting, in my opinion, in making some of those challenges obvious. It's nice to have something that is based on evidence that, in a way, confirms some of the feelings that we in the technology field already had. One important thing is the algorithms of major companies, and how they select content and how they choose which videos or which content young people hear and see.

So, to move on to another related field, I also work on a project called Mapio Cymru, or Mapping Wales, that has the aim of creating mapping and geography infrastructure in Welsh, with funding from the Welsh Government. I thank them for that. The hope is that that infrastructure will then improve mapping services in the future in a variety of fields: leisure, satellite navigation, fitness and things like that. There's quite a bit of work that needs to be done in order to achieve that dream of having a sat nav app that would work completely in Welsh. I know that some people have already referred to that in the first session. I'm happy to expand on that project, and also I'm happy to talk about artificial intelligence. I define myself as a sceptic when it comes to some of the assertions that are made about AI, particularly those that come from the American corporate industry, but certainly there are things there that are relevant to this discussion, in particular voice-to-text, or speech-to-text, which was discussed in the earlier session, I know. 

I also have some viewpoints on the essentials, like dictionaries and spell-checkers, grammar checkers and things like that, in terms of helping people to use the Welsh language confidently. Also, another area is the opportunity to use the Welsh language in a leisure way, for example games for teenagers. So, that was my five minutes of themes.

Excellent. You've covered several different elements there. I'm sure we'll go into detail on those. So, thank you. That was really useful for us as an overview. Given some of the points that you've made already, we'll go straight to Mick, if that's okay.

Well, there is a great deal of important work outlined in the plan, and that's encouraging. I think it's excellent that the Government has focused on technologies under open licences that enable and hopefully encourage reuse of some of those resources. I think that the work on the essentials—. Llion Jones mentioned that Ap Geiriaduron, the dictionary app, for example, is very important and perhaps doesn't generate a great deal of excitement, but it's important in encouraging people to improve their vocabulary and their ability to speak Welsh and to speak Welsh more confidently.

In terms of the work as a whole, I would question how effective any efforts were to try to introduce services and content on the ground. So, there are common themes in terms of creating things, ensuring that things are available, and there is a job of work to be done then that is just as difficult to try to ensure that those devices are in people's hands. Gruffudd Prys referred to this as well in the first session—the fact that the grammar checker is available as a technology, as a software package, with training data that is dependable, and isn't based on artificial intelligence, but on the rules of grammar in the Welsh language. Is that reaching people who are on their phones or on their computers, who use different devices? That's a challenge and is a difficult task to achieve. So, I think there is a great deal of potential that hasn't been fulfilled yet in that regard.

When it comes to content, I was very pleased to see that there has been investment in Wikipedia through the medium of Welsh. I don't know if you're aware, but, when I run sessions with people, I ask them, what is the—I've given you the answer already now—most popular Welsh-language website in terms of visits? Some people say the BBC, and some people say different things, such as S4C. Well, it's the Wikipedia through the medium of Welsh. That's the most popular in terms of visits. We don't often see the word 'million' when it comes to the Welsh language—not yet anyway—but I was looking at the recent statistics and there are some 10 million visits to Wikipedia every month through the medium of Welsh. A subject researcher or a reader would, perhaps, visit Wikipedia 10 times a year, so that counts as 10 different visits. Wikipedia is a resource that is reused in several different contexts.

So, I wouldn't argue with the work that has been done. There are many important elements to that work. From my point of view, as a specialist working in the field, my interest isn't just the potential, but delivering upon that potential, and use of all kinds, and use by companies and providers, too. So, that is more of a question rather than an answer, because I don't have the statistics in terms of usage, as it were

As someone who works in the area and concerned about the Welsh language, I'm aware of the Government's major target to create a million Welsh speakers, who will also hopefully use the Welsh language day to day. There are always things that we can find—good projects that are deserving or opportunities that need to be taken—that deserve to be funded in order to buy that time for specialists to develop the technology.

There's one quite major area in my view, and that is the need to create experiences or activities that people can do through the medium of Welsh that have an element of technology related to them. For example, I'm aware that Ffit Cymru have done some amount of work in this area, but there is not much use of technology in Welsh in the fitness sphere. The commissioner's research highlighted that sports and fitness was an important element in terms of using the Welsh language in communities, and was quite an accessible way for children to use the Welsh language. There are things like dating apps that are very popular in other languages but don't exist in Welsh. But activities more generally, I think—.

There is a policy paper by Cymdeithas yr Iaith that talks about the idea of creating a Welsh digital initiative body, not just to create content and videos—that is already done to a certain extent by Hansh and some other providers like Golwg—but to create the apps, to create those experiences and to create those activities that then create opportunities to have fun through the medium of Welsh. I think there's a scarcity, if I may say so, of fun to be had in Welsh when it comes to technology. There are some games that are available through the medium of Welsh, and there's one person in Caerphilly who translates quite a few of them, who used to work for Menter Caerffili. He's called Morgan—that's his first name. He has done quite a lot of—. He has got in touch with companies and has worked with them to translate games into Welsh on a voluntary basis in many cases.

It would be nice to try and formalise this and to create more games that are available in Welsh. Perhaps that model would be quite similar to S4C, where there is the commissioning of brand-new programmes and there is also the adaptation of cartoons and programmes and things like that, to extend the plurality there. We don't see that plurality of content and activities that's needed in order for people to live their lives through the medium of Welsh, I would argue. Games can be quite expensive things. I don't work in the field of computer games specifically, but they can be expensive, but not necessarily either.

Well, open innovation—. The word 'open' can be quite slippery and ambiguous, in my view, but at its heart it refers to the ability to re-use other people's work and to collaborate on a global level with other people. I can give you one or two examples. The most popular website system is WordPress. If you want to create a website, you can speak to a developer to generate a site on WordPress. The fact that that is available through the medium of Welsh emanates from the fact that all of the software is available under free licence to copy. It's as if Coca-Cola had shared the formula for Coca-Cola so that people could make it themselves. That's at the root of this—the software. As it happens, it works commercially for the company behind WordPress to distribute all of the software, all of the code behind it, because otherwise the software wouldn't have been available in many languages. Developers then create add-ons, so there is a development ecosystem. I hope that I've described that accurately there.

There is a danger in the traditional way of proprietary software, where one company controls the software and then can tie people into the software. It's not easy then for third parties or us here in Wales to create our own adaptations or to create add-ons. We have our own needs in terms of the Welsh language, and the things that we want to do. So, software under open licence provides greater freedom for us as users and consumers. So, that's exceptionally important. That's the ideal, in a way. If you know anyone who uses or you use an Android phone, Android is based on open source and a free licence. It's mainstream in the industry now that companies release software under an open licence.

Thank you for that. Mick, are you happy for us to move on? Great, we will—

Thank you, Mick. We'll move to Alun. 

Thank you very much. You said, I think that it's fair to say, that you are a sceptic when it comes to artificial intelligence. Could you explain what you meant in saying that?

I'm a sceptic about the claims made by the industry, because we've seen many instances of technological hype in our lives, such as when the web was new in the 1990s. There was the dot.com bubble, for example. So, over the years we've heard about blockchain, cryptocurrency, metaverse, even cars that drive themselves—that has died down a bit recently. There is usage—there are thousands of cities in the world, and about 10 of them have adopted cars that drive themselves. What I'm trying to say is that there is an obvious advantage for corporations and companies to make baseless claims, and use hype to not only sell technology, but to sell shares and to try and create profit in that way. So, you have to be sceptical, but—. For example, I don't see that AI is going to take a lot of jobs; I just don't see the basis behind that story. Machine translation hasn't taken many translators' jobs yet—there is some sort of precedent there—but it has made the process of translating more rapid. They, as translators, are—. For example, machine translation is an example of AI, isn't it? They are a lot more efficient as translators; they can complete more tasks, so the AI is a tool, then, that facilitates them in doing that.

Well, it has changed the way that we work here, and I think that many people who used to work in call centres would disagree with your analysis about job losses and so on. We have seen many jobs being lost, and that's happening on an ongoing basis. But perhaps the speed of travel doesn't always reflect the kind of discussions that we have. But I think the direction of travel is quite definite for the future, namely a greater use of AI than we had imagined in the past. Do you see that there are opportunities for the Welsh language through greater use of AI?

Certainly, there are opportunities; there are hazards too. The main opportunity that I perceive is to help people or groups or any entity who wants to create things in Welsh, whether that's a video, whether it's audio. It arms people, it gives people the tools, to create more food, as it were, for the internet. So, when we look at the Welsh Language Commissioner's research, for example, we can see the references there to these major video platforms like TikTok, Instagram. It would be interesting to see more research on this. But there's a lack of provision and a lack of plurality in terms of the material that exists in Welsh.

One technology that is transformative, I would argue, is subtitles—. Well, the ability to take a voice recording and to turn that voice recording into text. I think that there is a scarcity of videos subtitled in Welsh. For example, when I was learning Welsh as an adult, I found that, on Clic at the time, S4C didn't provide many Welsh language subtitles at the time. I don't know what deaf people do. I imagine that they read the English and take things in that way, which is far from ideal, I would say.

Speech-to-text then allows people—. There are very experimental systems now where we could do a video about a community, for example, and then edit the video by editing the transcript, which would then be a much easier way of editing a video. So, there are a lot of opportunities there, I think.

So, what are the risks?

In terms of use of AI?

Yes.

The first thing I would like to say is climate change, and the energy that AI and the generative models use. That is enormous.

Yes, but that isn't related to the Welsh language, though, is it; that's about the industry as a whole, isn't it?

I think it does affect the Welsh language. In a report by the Commission for Welsh-speaking Communities, for example, there is a section there about climate change and the relationship between climate change and the Welsh language. I think—. I don't think there's an easy solution or answer to this, but—. There's a slogan that campaigners use: 'There's no Welsh on a dead planet'. We need to take that challenge seriously, I think, especially before we reach net zero.

Yes. I understand that argument, but I'm talking about the Welsh language and AI. You said that there was a risk there, or 'hazards', I think, was the word that you used. So, what are those hazards? Can you describe them, for the Welsh language?

I think there's a risk as well in terms of providers that don't share the same objectives that we have here in Wales, who think they can just add the Welsh language in very simply. There is one recent example—Duolingo, for example. I understand that Duolingo are not updating the Welsh language course at the moment. But they turned to an AI-first policy recently, which is quite a silly decision, in my opinion, to get rid of those language experts and to try and do everything with AI. But this is just a sort of example of what companies can do. We could see a multitude of content available on the internet, and a lot of it would be slop, in Welsh that had been generated automatically, of very low quality, and that might perhaps drown out the important things that we want to publish as the people of Wales.

Yes, but there's a lot of nonsense in English on the web as well. That's the kind of world that we are living in at the moment, and, if you generate more content, well, not everybody wants to see all of that different content. So, in principle, providing or creating more content through the medium of Welsh is a good thing, I would imagine. That doesn't mean that I want to see or read or watch every kind of content created. But that's true in English as well. That's natural, surely.

Exactly. There is a lot of nonsense in English. There are a variety of opinions as to what exactly is nonsense, from one person to another. The difference with English is the investment in quality content. That quality content does find its way then, and people can recommend and share quality content. There are many providers. But there isn't one designated body in Wales that has responsibility when it comes to creating these experiences and this content on the web and that is entirely committed to the web. I understand that there's a lot of work being delivered by the tv industry, the publishers, but the internet is always just one element of the work that they do, if you like.

Nobody wants one body that generates all of the content, do they?

Not one that makes all the content, but that—

I don't want a body creating my content. I have my own content to share that I provide on the web, and people watch it.

I think that Heledd wants to come in with a question.

I take it that it's more about the standard of the language rather than the content; that's the point that you're trying to make, isn't it, in terms of, if it's generated by AI, that there is a risk in terms of the standard of the language that people see as well.

I think so, because especially in the Welsh language learning sphere—. I've always had doubts about Duolingo, for example. I also think that those job losses are a big danger, because, in some cases, employers are jumping into AI very early in the process, and based perhaps on the hyperbole that they are seeing and hearing from the industry. Is there anyone in this room who wants to speak with a robot rather than a person? That isn't close to the ideal in terms of good service, even if it is believable as a synthetic automated voice. It's about the sort of world we want to create and the sort of Wales we want to create as well. I know that there is no easy solution in terms of forcing some companies to do this at the moment.

May I ask a further question? You mentioned pluralism, and you referred at the beginning to the report of the Welsh Language Commissioner published last week with regard to children and young people. That was one of the things that came through very clearly, especially when it comes to young people, once they're in their teens. Perhaps the content for younger children is there, with Cyw and so on, but when it comes to the social media platforms, and you also mentioned the algorithms and the kind of content provided, do you have any recommendations when it comes to how we could overcome those particular barriers? Who should be leading on that kind of work? Because clearly with pluralism, you referred to S4C, and there has been a role for the UK Government with the young audiences fund in the past, and the Welsh Government also has invested to ensure that content can be provided. Are there further steps that you think we should be taking in this area in particular?

Well, the corporations have a lot of power. When we look at what young people are doing and the research from Iaith and the commissioner, for example, there is one Chinese company, TikTok, and the rest, I think, in terms of those big players, are American companies. That's the challenge. If they were in this room with us today, if TikTok were here in the room, I would want to discuss the algorithm with them and how Welsh language content can find its way to people who are going to appreciate that content. Because at the moment, the way that they select content does favour things in the major languages—English in our context here—because of the numbers of people. In a way, there isn't much hope that a creator will reach millions of people and reach that premier league of content creators. I see that there is an opportunity here, perhaps, to work with other Governments to create a coalition. You spoke about this in the earlier session in terms of opportunities to collaborate. Well, what about some sort of coalition of countries, a coalition of global languages that would then influence these corporations? They are very powerful in terms of what children see from day to day. It is they who create that linguistic environment, in terms of the use of mobile phones, and so on. 

There were three main reasons that the respondents gave in the commissioner's research: 'There is not much Welsh on the platforms that I use.' That's true for adults as well. If you go on LinkedIn, how many people use the Welsh language there? How many people use Welsh first? It doesn't feel like that many. I don't have the official statistics in front of me. One of the other reasons was, 'Not enough confidence to use the Welsh language on platforms.' Confidence is interesting, isn't it, because how can we give people more confidence? That is very complicated. I certainly think that there is a role for some devices in terms of spellcheckers, grammar checkers. I don't want to say that there's an easy solution to this, but there are ways in which we can provide a little bit of reassurance to people who want to use the Welsh language online, who have that goodwill towards the Welsh language, to help them to use standard Welsh. I personally don't mind one way or another when it comes to less formal Welsh, but for some people that is important, and the provision of devices or software in the right place is an important element in terms of—I've forgotten the term now for 'predictive text' in Welsh. But with something like predictive text, which is so vital to everyone's use, but especially young people, of the Welsh language, there are some systems, but they haven't reached the grass-roots level yet. Those essentials are going to play an important role.

Again, a lot of this stuff is available, either through Welsh Government grants or volunteers, even. Android was all translated by a volunteer in Rhosllannerchrugog, Aled Powell. It's available, it's out there, and if you're a geek, you can set Android on your phone to Welsh, but most of us aren't geeks, we just want to use things by default.

What is iPhone in Welsh?

Things such as auto captions aren't available on all of the platforms in Welsh—

Certainly.

—as an option. So, that makes it even more difficult, doesn't it, to get that content out there with subtitles in Welsh.

Certainly. I think there is a major cohort of people who are going to benefit from having that content with Welsh language subtitles, because perhaps there are some dialects that might not be familiar to them, but they've been through Welsh-medium schools and they just want a little bit of help to understand the content fully, and then they become more familiar with the different dialects and so on. So, on Clic, for example, I think that the software does support Welsh subtitles, but for some reason S4C at the moment isn't investing in Welsh language subtitles as far as I can see. I hope that the developments that have been shared by Bangor University and others, perhaps, later on, will contribute to decreasing the cost of doing that.

Thank you.

Okay, thank you. We'll go to Gareth.

It's interesting, there are thousands of apps and it's a very small proportion that offer a Welsh interface. I would tend to think that there is more provision in the educational category than others. I've mentioned this when talking about fitness apps for example, like Strava, which ties use of technology with the outdoors and keeping fit. We can see that the Welsh dictionaries app, for example, has 50,000 downloads, which perhaps demonstrates that the users are out there who are willing to use things like this. Apple doesn't publish figures, but usually it's possible to get some sort of estimate in terms of the figures on Apple based on the figures that Android do publish.

Another thing I should say in terms of the development of technology is the ability to create websites that behave like apps by now. There are technologies like HTML5, for example, and progressive web apps. When I work with organisations, I have to remind them sometimes that they can create things that feel like apps within a website. In all cases—. It depends on the requirements, but you don't have to create an app for iPhone, an app for Android that are separate in every case. That's not always necessary. One interesting example is Wordle, for example. Wordle was a game that became popular very rapidly through a website app online, and it capitalised on some developments in the background in terms of the ability to do things within a web browser. There are games now that run within web browsers. So, I perceive that as a big opportunity for people who want to generate activities and experiences.

I think that that's a really good question, and all of the questions asked have been excellent, by the way. I think it then leads back to the power of the corporations and the major companies, and how we, as a nation or as a group of nations that wish to revive our languages, can influence these corporations. There are things, there is a precedent for this. The first desktop word processor, for example, that was available through the medium of Welsh was OpenOffice, and then major companies and providers such as Microsoft realised that this was an important category, or a set of important functions. So, it starts with the world of geeks, very often, and the amateur sphere, the volunteers, where people create Welsh versions of things. There are a lot of things that are available through the medium of Welsh and the challenge is to use the power of Government to collaborate with those corporations. Historically, Microsoft has been relatively good and they acknowledge the Welsh language as an official language. Apple hasn't been quite as good historically. There is variation. Google is somewhere in the middle when it comes to awareness of the need to support the Welsh language.

I don't personally see any technical barriers, as such, to this. There are banking apps that are multilingual already in existence. The lack of provision is because of a lack of will on their part, or a lack of enforcement by the authorities in Wales, I would argue. What was HSBC's slogan? 'The world's local bank', wasn't it? They need to be challenged on that point.

Yes. We made that point as a committee. I agree.

There's an important role for authorities in Wales to play in terms of the standards, to put it bluntly, in my view. As someone who works in the field of technology, I don't think that there is a technological barrier in terms of what could be provided. Any bank that would say that these sorts of barriers exist would be just using a baseless excuse, in my view. 

That's really useful. Thanks for that, Gareth. Moving back now to Heledd. 

Thank you, Chair. You've spoken a great deal about the major technology companies and, obviously, the huge power that they have. I was very interested in what you said about the possibility of forming alliances or coalitions with other countries and other languages. Are there things that you think that we should be focusing on from the perspective of this Senedd, the Welsh Government in particular—areas—whilst trying to forge those coalitions, those key areas that we should be focused on in trying to gain greater influence and a greater profile for the Welsh language, such as AI? Or do you see this as being pertinent across all of the areas when it comes to that target of a million Welsh speakers, but also increasing the use of the Welsh language?

There are several areas. I think that AI is part of the discussion with them, in particular when it comes to the ability to provide things like subtitles through the medium of Welsh and other things that we have already discussed. It's the algorithms that are the challenge, that internal formula that those companies use to provide content. So, I was discussing with someone in the field recently, an academic who was researching TikTok. And it's very difficult to do research into TikTok, because every user gets a completely different experience of it, or perhaps not completely unique, but relatively unique in their experience of TikTok. So, yes, if they were in the room, I would want to discuss algorithms with them. I think that Welsh interfaces are important as well—that ability to just click on buttons that are all in Welsh. To what extent? I'm not entirely sure.

I like to quote an expert from the Basque country, Luistxo Fernandez: 'A menu is important, but we need food in our language'. It's nice having that Welsh language interface, and the ability to click the file, and edit and all these different buttons, but we need that substance, that content, that food, and that food for the menu, and we need to find a way for users to be able to see and hear that content. Predictive text—I've remembered what 'predictive text' is in Welsh—is incredibly important, I think, in terms of the commissioner's research—those fundamental things that just give a little bit more help for people to type in Welsh. Have you ever had that experience where you're typing something in Welsh and you type the lowercase 'i' and it turns it into a capital 'I' automatically because it assumes that you're typing in English? It's things like that that are a major barrier for a lot of people on the ground, I would argue. And we can then increase, perhaps slowly, but we can increase the use by providing things of that kind.

Thank you. You've mentioned several times the importance of research, and you referred to the importance of the report by the Welsh Language Commissioner, because it's easy to follow assumptions. It's important to get that confirmation about those things too. How important do you think it is that there is more research in this area, so that we understand better the use of technology by Welsh speakers and new Welsh speakers?

I think it's hugely important , because research is undertaken very often in other languages. There is a great deal of research about the use of the English language on the web, for example. Perhaps it isn't labelled as 'use of the English language', but several experts who speak English would look at use of the English language without questioning that. But I think we need to research in terms of the content that's produced, to what extent people, young people in particular, receive that plurality of content. I would argue that there isn't that plurality of content, but I'm saying that based on instinct. I don't have the formal statistics. I'm aware that Ofcom is already undertaking research where there is monitoring of devices with the permission of participants, of course—so, researching those people who keep journals of what they do online. We need to interview people to learn more about their habits and practices and how we can empower them and equip them to overcome some of these barriers. It would be good to have that dependable evidence base, because a lot of this is instinctive or on the basis of personal experience. 

We need to look at successes in different languages. There is an example in the Basque Country where they've created a Wikipedia for young people between the ages of eight and 13 called Txikipedia—I think that's the pronunciation. It draws on the work that's been done on the main Wikipedia site, but it's aimed at people in that relatively specific age range of eight to 13. We should look at what they've done, to see whether we can emulate the work that they've done there. They've also tailored Txikipedia to respond to some of the needs of their curriculum, for example, across several subject areas. So, perhaps that's an opportunity to collaborate with other languages as well.

Thank you. Could I just ask you one other question? Evidently, you've spoken about the algorithms and how that is something that is very complex to solve. Obviously, we would hope there would be ways of doing that. Do you think that there would be a role, potentially, for the Government or some other body in Wales in terms of promoting what is available? Because very often, the work that you've been been engaged in, and a number of other people, where there are resources available through the medium of Welsh, and there is content available, but it's almost the case, because of the way that the algorithms work, that people won't come across that content through the medium of Welsh. Do you think that there's a promotional role needed in terms of what's available and how people can live their digital lives through the medium of Welsh, or is that something that isn't relevant to Government?

Yes, I do think that's the case. I have conducted several very informal surveys in terms of specific content. At one time, I ran the Hacio'r Iaith conferences over a few years, and I was surprised, even for people who spent a lot of time on the web and online, I was surprised to see that some content wasn't reaching them, when it was things that some people would consider to be the best content that's available.

Promotion is one way of thinking about it. I think there are a lot of ways of thinking about it in terms of how people and content can come together. Are we then trying to get more value out of what already exists as well? There's a role for marketing to play in terms of promotion, but I also think that there's a role for other ways of thinking about the challenge. Perhaps we need to appoint a group of people who would go onto the platforms and influence them.

Just to jump to something a little bit different, I also think that there are opportunities for us to create Welsh platforms. We can't deny that Instagram and TikTok is popular, but there is space for specific platforms that we could create in Wales, perhaps jointly with other languages, that would be closer to that aim that we want to achieve as Wales, if that makes sense. We could try and knock on the doors of Instagram and TikTok, and we could try and talk about the algorithm—and we could do that for years to come—but it isn't likely that we could create something as big as Instagram in Wales at the moment in terms of the current factors that are relevant. I'm not saying that we need to try and recreate the way that they operate, or to recreate the size of those companies or anything like that, but just to think a little bit differently.  Is there room for a completely Welsh language platform that would be leisure related, or a number of platforms that would create that space? That idea of a space or immersion in the Welsh language is quite a consistent theme, I think, from the Mudiad Meithrin and the cylchoedd meithrin to spaces like the National Eisteddfod, Eisteddfod yr Urdd and other events. That idea is just as true in terms of the internet as regards trying to create more spaces where the Welsh language is hegemonic.

Great. Thank you. A good word to end on there—'hegemonic'. Excellent.

Thank you for all of the questions and thank you for the challenges. I appreciate every one.

We've really appreciated the evidence you've given. A transcript of what has been said will be sent to you to check for accuracy. I'm sorry that we've run out of time, but thank you so much. It's such an interest subject. This has enriched our understanding of, as you said, some of the challenges that we have to be aware of. So, thank you so much for your evidence this morning and into the afternoon.

That's one opportunity to use technology, isn't it—speech to text—to accelerate the job for a colleague who works here on transcription, fair play to them.

Excellent. Thank you very much.

Thank you.

Members, we will take a lunch break now. If we can be back by 13:25, that would be very useful. But, Mick, I know that you won't be able to join us then. Thank you very much.

Welcome back, and good afternoon. We're now moving to item 4, and continuing with our inquiry into Cymraeg for all, and an additional evidence session now. Before we get started, I'll say that we have received apologies from Mick Antoniw for this session. So, we will go straight on to questions. We have three witnesses with us this afternoon. I'll ask them to introduce themselves for the record, and I'll go to Dr Mac Parthaláin first of all to introduce himself for the record.

My name is Neil Mac Parthaláin. I'm a senior lecturer with Aberystwyth University. I am also a researcher into artificial intelligence.

Good afternoon. I'm Dr Cynog Prys. I'm a senior lecturer in sociology and social policy at Bangor University, with an interest in the Welsh language in the digital sphere.

Great. Welcome to the three of you. Not every one of you has to answer every question this afternoon, but we very much look forward to hearing your evidence, and we'll go straight to the questions if that's okay. If you could, at the outset, just give us an overview of how you perceive the digital landscape, where the Welsh language is in that regard, if we're in the vanguard as compared to other languages or not. How do you perceive our current position? Dr Prys, you nodded your head, so I'll go to you first.

Thank you very much. In terms of the Welsh language, I think that we have been very innovative in Wales, and generally I think that the situation of the Welsh language in the digital sphere is very positive, and it has been so for a number of years. I won't list everything that we have in Welsh now—I'm sure that people in this room already know of some of the technologies that we have—but I think what I'll do is try and pick out some of the most significant things that we have.

I would perhaps start with the recent developments in generative AI. I think that the quality of something like ChatGPT in Welsh speaks volumes about the success of the technological strategy. It is incredibly good and it seems to be improving every time I go on it, in my view anyway, in terms of the quality and the accuracy of the Welsh language. It opens all kinds of doors to exciting and new opportunities in terms of the Welsh language. I think that shows where we are in terms of the level of success in the digital sphere. I think the other technology to discuss is the speech-to-text work, and the fact that Microsoft Teams is now available in Welsh. That is quite a high-level thing to have available for a minority language. That's just come out and the quality is fantastic.

So, if someone asks me where are we in terms of the Welsh language, I would say as a minority language in Europe, we're in the upper half of the table compared to other languages. We are behind the Catalans, which is a much bigger language. We're behind the Basques, who have more resources as well, but I think the Welsh language sits somewhere behind those two. So, I think that the general overview I would give is that it is positive, and we just need to keep going with that momentum we have at the moment.

Thank you for that. Did anybody else want to come in on that point? Dr Cunliffe.

Thank you for that. Dr Mac Parthaláin, was there anything you wanted to say on this first question in terms of that landscape and how you perceive things?

I find myself agreeing and disagreeing with some of the things that have been said already. One thing that is perhaps a cause of concern for me is the report—there were several articles on the BBC about this recently—about the fact that almost all of the language that is written by first-language Welsh speakers online is in English.

I think that that perhaps dates back to the time when Government policies weren't quite so much in the forefront when it comes to the digital landscape. Everyone remembers, when the internet came, that there weren't policies that had been formed for anything that was happening online.

I think, in those terms, that the Welsh language has lost out digitally. I hasten to say that there's a better attitude towards it these days, and I think that everyone is agreed on that, that there is a better perspective available on that now, and that everyone knows that things like generative artificial intelligence will be here—whether we like it or not.

But I would disagree with some of the points that have been raised already. I don't think, especially if you look at the younger cohorts, that young people have the same attitude in terms of using the Welsh language online in a way that is natural, if you like.

Thank you. It's important and useful for us to have a range of views, so that's very useful. We'll move on now to Heledd.

Thank you very much, Chair. Dr Prys, you mentioned some of those things where we have been in the vanguard. Clearly, in terms of the Welsh language technology action plan between 2018 and 2024, it was about ensuring that Welsh was at the heart of user experience. To what extent do you believe that that action plan has been successful? And I'll open that out to the others as well.

I'm going to give a positive answer first of all, and then I'm going to find some blame to place after that, rather than just sticking to the positives.

I think there's a lot to be lauded within the strategy. I think something that's interesting and something that has borne fruit is the fact that the Welsh Government have a policy of making their technology open source and open to technology companies to come and to use and to adopt that technology within their models, without even having to say that they have done that to anyone—they can just go and use it.

That has allowed major commercial companies, where there isn't a statutory requirement for them to do stuff through the medium of Welsh, to come and to use Welsh Government-funded things, and things they've put up on GitHub for them to use. I think that that has been a way of influence.

I know that there has been lobbying going on by the Welsh Government as well. They've worked with these major tech companies to do things like Microsoft in Welsh and Google Classroom, and all these kinds of OpenAI things as well—all of that group. So, I think that those elements have been wise. I think they've put their money in the correct places. I won't give too much of an opinion as to whether they've given enough money—that is another issue.

But I think that one piece that's missing from the strategy puzzle is that we don't know enough about how people use technology, and I'm here as a sociologist rather than as a pure tech person. I think Daniel touched on this: we know things about social media. That is a theme across Europe.

Minority languages have had research done into social media and the use by young people of Facebook and so on. We know a little bit about that, and we know that people tend to use the majority language, so it isn't a great picture. But what we don't know half enough about is how people in the educational sphere use technology, for example, how people in the workplace use this technology.

If we think about where these apps are used, where this technology is used, that educational domain is incredibly important, and the workplace in the bilingual modern Wales is incredibly important. But we know barely anything about how people make use of these things and what people want in Welsh. Do we know what people truly want, or are we just guessing?

So, that piece is completely missing from the strategy, but those things that they're talking about, the things that they've funded in terms of specific technologies and technology spheres, I think that those are good things. But, yes, we need to know a lot more, just speaking as a sociologist.

Dr Mac Parthaláin, was there anything you wanted to add on this?

No, no comments to add. I agree with the points made, to be honest.

Thank you very much for that. Just in terms of the new priorities for Welsh language technology that have been outlined, namely technology to increase the daily use of the Welsh language, making sure that everyone can access Welsh-language technology, improving Welsh language artificial intelligence and speech and language technologies, are they the correct priorities, those three most important things? Is there anything else that you would have liked to have seen prioritised? I know that you didn't want to note the budgetary element, Dr Prys, but clearly in terms of the investment that's needed, do you think that these are the priorities that are needed and that the funding is in place? Anyone else is welcome to contribute too.

If I can go first. I think that they're areas that need to be looked at. I think, to pick up a little bit on the point I was making earlier, we need to ensure, as you just said there, that this technology gets into people's hands, the hands of those users out there. And one thing I'd like to raise is to ask a question: whose work is it to promote that this technology is available and to advertise that? I think that we have more language technology than we have awareness of that technology. 

I have an interesting example as a lecturer at a university. I have asked, in these past few years, to students, 'Do you use predictive text in Welsh on your phone when you text or write an e-mail?' And none of the Welsh speakers I lecture to—none of them—use predictive text. They don't know that SwiftKey and DBoard exist, and these are first-language Welsh speakers, most of them, who have been taught in Welsh-medium schools. How have these people gone through the education system without finding out that you can—? What they do is they turn off the predictive text to write in Welsh.

There's something missing here, isn't there? So, to get this technology into people's hands, we need to add another element to this strategy. So, we need technology champions, perhaps; we need individuals to go out there into the schools, when they get their smartphones for the first time when they're 13 or 14. What about going into the school and saying, 'Did you know this existed? Did you know that you can do this in Welsh? We have this technology.' Predictive text has existed in Welsh for many years now. It's such a shame that we have this technology, but that people don’t know that. The fact that Teams can transcribe in Welsh now—. I just stumbled upon that whilst speaking in an event, and I saw suddenly that it was transcribing my Welsh. I almost stopped speaking because it was quite distracting. [Laughter.] But it’s incredibly good at that. Who goes out into the workplaces? Who goes out into the schools? Who works with the teachers to tell them, 'It’s possible for you to change the language on Google Classroom into Welsh'?

In that regard, clearly, we as a Senedd did just pass the Welsh Language and Education (Wales) Bill, and promoting the use of the Welsh language is part of that Bill too. So, the argument, therefore, is perhaps we need to have that research to understand better how people are using these different routes. But it’s that promotional element of what we already have that’s currently missing.

Certainly. I think we have a lot of resources, we have a lot of equipment, but I question how many people know about these things. And that, then, interferes with people’s ability to use this technology. And for people like the Welsh Government to get the full success after doing all the right things that they’ve done, that last step is missing, I think.

You were nodding there, Dr Cunliffe. 

Thank you. Dr Mac Parthaláin.

I would agree with a great deal of what has been said.

One other thing that I come across, particularly amongst students, is the tendency not to follow a path where the Welsh—. Even where the student themselves is a first-language Welsh speaker, they don't pursue an academic route through the medium of Welsh, although there are many opportunities available in that direction. And I think that one thing that isn't the focus of this group, but perhaps has some—how can I put it—other focus, is this idea of how wide-ranging and widespread the use of Welsh is amongst private corporations and public sector organisations as well.

Many of the final-year students that I see say, 'Well, what's the point in doing a project through the medium of Welsh, because after I leave I won't be able to use that language in my work?' and so on, and it's very difficult to explain to them that there are many posts out there where the Welsh isn't just desirable, but is essential, in some areas. I think that that link is missing—the link between use of the Welsh language and the fact that it can be used in a professional context as well. Perhaps that isn't a problem in other areas, in terms of the media and so on, where there is a clear link there between the profession and use of the Welsh language, but I think, particularly when it comes to areas such as the sciences, there is no link made there between the use of the Welsh language and studying through the medium of Welsh, and then using the language professionally. I am aware of the fact that that isn't the focus of this current enquiry or group, but perhaps some work could be done in that regard to strengthen that idea that there are jobs out there through the medium of Welsh and that there are careers available using Welsh in the field of technology.

We'll go to Alun.

Thank you very much for the evidence. I have enjoyed this discussion. Several of you have said that it's difficult to imagine where AI is going to go, so I'm going to ask you to imagine where AI is going to go, and how can the Welsh language be part of that?

Right, I'm going to issue a disclaimer here that I'm a sociologist, so you're not going to have a technical response from me. We are looking at social change here. As sociologists, that's one of the things that we do, and we've done that since the beginning. So, the direction in which this is going is that the technology is going to get better, and AI is going to be part of everything that we do—every public service that we use will have AI running in the background, businesses will use AI in the background. You won't be able to speak to people and say, 'Well, no, I don't use AI.' That isn't going to exist, because everybody is going to use AI in some form, whether they know it or not.

In terms of the Welsh language, there's an opportunity here, because I think what will be very good with these models is, I think, increasingly, they will be included and incorporated in public services online in terms of chatbots, as part of the services, perhaps, that we receive from people at the moment, and I think increasingly, we'll be moving towards—. You have a question about your recycling bin, you'll contact your local council, and perhaps you won't speak to someone in person—because it's cheaper, basically, to ask a chatbot to respond to the question—you'll ask your computer and a virtual person, perhaps, will respond to your question. That technology already exists.

What's interesting in terms of the Welsh language is that it will be possible to make this work in Welsh, even orally, so that you ask your digital assistant in the house, in Welsh, 'When will the bins be collected this week?' That will happen. I see Welsh language services being provided not directly by people, but through computers; those will be easier to provide through the medium of Welsh in future.

It does raise questions about private businesses that don't come under the auspices of the legislation at the moment, and people are reluctant to extend the standards to other bodies, because of the associated costs. Well, I think that the translation could be cheaper, and that there are ways to work around these issues so that services are provided through the medium of Welsh in a variety of ways.

So, that's going to be interesting, to see the possibilities there, but I think, just going back to my previous statement that we need further research on this: do people, and do Welsh speakers specifically, do they want to receive services through the medium of Welsh from a computer? We've heard a great deal about local banks closing down in communities the length and breadth of Wales. You used to be able to go into the bank branch to ask for a loan, to ask for assistance, and you could do that in Welsh. Even the phone lines, the Welsh language phone lines, have gone now. Do people want to have a chatbot saying, 'Well, no, you can't have a loan from us'? Do people want the Welsh language provided in that way? It's something that's possible to do, it's achievable, but we need to know a great deal more about what people want and what their aspirations are in terms of technology and the Welsh language.

I don't know whether I've answered your question, but—. I see that there are opportunities; things will improve, but there are huge questions about what people really want.

I want the HSBC app in Welsh.

So do I.

I don't want to a call line—I know that we've discussed this, but I've never phoned them. What I want is to be able to have the app on my phone and to be able to live my life in Welsh in the same way I can in English. I take it that that's possible today.

There's no reason at all why that wouldn't be possible. They could easily do that.

Dr Prys wants to come in, but, just before we do that, I know that Dr Mac Parthaláin wanted to come in on what Dr Cunliffe was saying, so perhaps it would be an idea to come back to Dr Prys afterwards, if that's okay. Dr Mac Parthaláin.

Well, yes, a lot of the things that Dr Cunliffe said there already exist, but the problem for me in terms of that might be that a lot of the scenarios he's explained are a bit unnatural, in a way. Yes, we can definitely do those things already—it's easy enough to do—but there are still problems in a technical sense in terms of accuracy and so on.

What would concern me most, to answer Mr Davies's question, in a way, is—. For us, even, who work in this area, we have been amazed by a lot of the things that we've seen in the past two or three years. It's difficult to keep up, even, with all these new developments in the field. A lot of them are driven by major companies, and major companies have all the power in terms of that. So, in that regard, it is possible to do these things; you'd need quite a lot of resource, I think. But where we're going in terms of these developments I can't say; I'm a bit afraid to say, because some of the things that I've seen, in particular in the past two years, perhaps 18 months, even, they've amazed us all in the computer science field and the AI field generally.

So, I wouldn't like to try and guess where we're going. I can certainly say that there will be developments that we won't like, as human beings, and I think that there will be a lot more automation and so on happening in some aspects and in some jobs, perhaps. Where that leaves the Welsh language I don't know. I think, in the end, it won't matter what language someone speaks, I'm afraid to say, because a lot of the things that happen under the surface will happen automatically, which would perhaps simplify life, in a sense, but I'm also concerned that there might be some detail lost. I'm starting to think about things like dialects and so on. The language might not perhaps be as—how can I put it—varied, or quite as—. It would become homogeneous, perhaps. That's what concerns me in terms of automating a lot of the things that Dr Cunliffe spoke about there.

Just one brief point at the end there, as one of the co-authors of that report. The opportunities are genuinely incredible, and would transform all kinds of things in terms of the way that we're going to live in future. But this ties back into the initial question, I think: what do people want? What do Welsh speakers want? We have to keep a weather eye on that question. How do we see and perceive the future of the Welsh language in our communities? I think, as we develop all of these excellent things, all of these exciting things and perhaps quite strange things, we need to bear that in mind. We need to speak more to people about these developments. 

It would be worth—. If we could circulate that report to Members, I'd appreciate that. 

Yes, definitely.

But Welsh-speaking Welsh people, like any linguistic, cultural, national community, they all want different things. I would like to see Cardiff City through the medium of Welsh. Even when I don't want to see what's going on, I'd like to see that through the medium of Welsh. I would like to be able to hear about Glamorgan tomorrow evening in Welsh. My son, who lives in Hay-on-Wye, right next to the border, I'd like him to have more opportunity to hear Welsh in everyday life, because when you grow up, having that opportunity to hear the Welsh language is an important thing. So, the community wants different things. What I want to hear is the what the opportunities are for us—as politicians, if you like—to ensure a space for the Welsh language in whatever developments may come. So, when my son plays on the Xbox, that that is in Welsh, and that the interface is completely naturally through the medium of Welsh. Whatever I want to do in English, that should be able to be done in Welsh, like having help to fill in tax forms through the medium of Welsh, rather than switching to English. Most of us do just switch to English because it's easier. We have become used to dealing with those things in English. So, I would like to know if there is a way, or will there be a way, to ensure that the Welsh language is a natural part of whatever developments we see in the future.

I'm going to ask you to answer that question, then, unfortunately, we're going to have to move on, because it feels ridiculous, but we are into the last quarter of an hour.

Oh, I'm sorry.

No, no, the discussion has been so rich, and there are so many additional things we've covered. So, sorry about this, but we will have to move on after that question. So, whoever wants to go first.

Very briefly, perhaps. The difficult thing there is to persuade tech companies to do things through the medium of Welsh—persuading your Microsoft Xbox, persuading these companies that are beyond legislation to do things in Welsh. I think that that is the challenge. But just a brief note on computer games and young people playing computer games: I think that that needs to be a priority area for the Welsh Government, to think about how to create a digital space online. The computer games industry is huge. I think that the expenditure there is greater than on film, if I remember rightly. How do you harness the enthusiasm of young people for that technology and dovetail that with the Welsh language?

I'll give a brief example before we move on. There's a group in the Rhymney valley, in Caerphilly, that is producing a YouTube channel called Yn Chwarae, where young adults come together and play computer games through the medium of Welsh. They stream it. These are people over the age of 18. They stream it, and that's an in-person space, because they come together to play these games, but they also stream it. So, there's a digital space where people can enjoy that technology—like your son, at the age of 15, perhaps, being able to enjoy and see role models, other people, doing this. So, it's not so strange to think, 'Well, I want to be a Welsh language influencer, I want to do things in Welsh.' It normalises it. So, that ticks a lot of boxes. I think that we need to do work in that sphere.

Thank you. We'll go to Dr Cunliffe, and then I promise I'll come to you too, Dr Mac Parthaláin.

Thank you for that. Dr Mac Parthaláin, you have been very patient.

Yes. Just to repeat what Cynog said there and to go back to the same point that I made earlier: there are opportunities available through the medium of Welsh, and I think that it's important for us to normalise, in a way, the fact that the Welsh language is a living language, and that it can be used in your career, and that it is an advantage in a way. I don't think that that is said enough. There isn't enough emphasis on the fact that there is a world outside education for the Welsh language professionally.

Thank you. We'll turn to Gareth.

Dr Mac Parthaláin, do you want to go first on this?

Okay. I think that a major factor in this is goodwill. I've noticed with many companies, if there is goodwill, or if there is experience of creating apps that are multilingual already, then you don't need to do much convincing for that company to provide through the medium of Welsh. That's my experience, anyway.

Some even—okay, I'm talking about websites here, rather than apps themselves, but some websites do translate from one language to another, if the website has the right plug-in. I remember something from Norway at one point where the interface was all in Welsh, and that surprised me because I couldn't work it out. What had happened, basically, was that there was a plug-in available on that website that had identified that I use the Welsh language, so everything appeared in Welsh for me.

So, if the will is there, I think that it's possible to provide a great deal through the medium of Welsh. In terms of measuring that, I don't know how one would go about measuring that, because it is a voluntary choice, in a way, in many cases. In other cases, people go out of their way, because they might be Welsh speakers themselves or whatever, to produce a Welsh version of their website or their app or whatever it might be. But, in other cases, it's a matter of goodwill, as I said. On the question of how you measure that, I don't know, to be honest with you.

Who wants to go next?

I'll answer in Welsh. I had a PhD student, now called Dr Siân Pritchard, who did a doctorate looking at exactly this topic. I can share that with you later. What she found in her work was that children's apps and educational apps were the most popular ones that people had on their phones. There were specific providers like S4C who created apps, and they were amongst the main providers in that area. Then, there were public sector apps, to a lesser extent. But, there wasn't much choice of Welsh language apps out there—yes, children and young people's apps, that's what they were.

Her users—she went out and she interviewed users in Gwynedd and conducted a survey with users in Gwynedd, and one area that they discussed was banking. That was an evident one. People would say, 'I want a banking app through the medium of Welsh. That's one of the main apps that I use in my everyday life. I used to be able to do that through the medium of Welsh.'

The choice is limited in terms of the apps that are available, but I think, with technology and the developments in technology, it would be possible to do a lot more. The app wouldn't necessarily have to be in Welsh, you can get some sort of system now that can just translate, as a sort of screen over everything, that can translate from any language into the language that you want. That technology already exists. So, perhaps it's more to do with the way that you set up your phone, or the settings of your phone. So, you would say, 'I want to read things in Welsh on my phone', and it should be possible to do it like that, rather than asking every provider to create content in Welsh. So, there are different ways of doing this, but we use apps every day, so it's an important area.

Thank you for that.

That's quite a wide-ranging question there. If I could ask for succinct comment on that question. Who wants to start off? Well, we haven't started with Dr Cunliffe yet, so we'll start with you.

Okay, thank you. We'll go to Dr Mac Parthaláin and then we'll go to Dr Prys finally. Dr Mac Parthaláin. 

I don't have much to add to that. I think that Dr. Cunliffe outlined the same comments that I was going to make. I find it difficult sometimes. I think a lot of these major companies—how can I say this?—they have one way of looking at what they do, and I think very often they don't see the demand, because so many Welsh people, Welsh speakers, speak English.

On the other hand, as a good example for you, I and the family went to Chester Zoo recently because we had seen a Welsh language advertisement on Facebook, in Welsh. So, there is a market. So, it's about trying to convince these people that there is a demand and that there is benefit, a financial benefit for them, on top of the fact that they're doing something good and that they are acting in an appropriate way and in a good way, if you like.

The final word to you, Dr Prys.

Why not be more ambitious? Why not legislate in this area? Why aren't we requiring it of these companies, who make a great deal of profit and who serve our Welsh-speaking communities. They're not going to not do that, they're not going to pull out of Wales if they're compelled to do this. Why not legislate?

That's a challenge. That's exciting.

To the battlefields. I was trying to translate that literally in my head to English. 

Thank you so much to all three of you. There are a number of things that we had hoped to ask, but the discussion was so extensive, as I said, we didn't manage to reach everything. If it's okay with you, we'll write to you with the additional questions that we have, and we will also send you a transcript for you to check that it's an accurate record of what's been said today. Thank you very much for being with us for an extra five minutes, and we are very grateful for your evidence. Thank you very much to all three of you.

You're welcome. Thank you. 

Thank you. And Members, we're moving straight on, but thank you once again to our witnesses, and thank you again to those online as well.

And Members, we're moving straight to papers to note. There are several papers to note in your packs. Are you content for us to note these publicly? Thank you very much. Everyone is happy to note those. 

So, I propose, under Standing Order 17.42, that we resolve to exclude the public from the remainder of today's meeting, if you are content to do so. Okay, we will wait to hear that we're in private session.