Latest / AI in the UK / AI in Mental Health and Education: Friend or Foe?
Transcript
- 0:02All right, your luck. Listen up.
- 0:18Hello and welcome to AI in the UK.
- 0:21I'm Neil Footman. And I'm Matt Wilkinson.
- 0:24Professionally, we help organisations, leaders and
- 0:27individuals learn about and implement this big scary
- 0:30behemoth called AI. There's some amazing stuff going
- 0:34on with AI in the UK that doesn't always get the attention
- 0:38it deserves. With AI in the UK, we want to
- 0:41change that. Hello and welcome to this week's
- 1:02AI in the UK. On this week's show, mental
- 1:06health issues are an increasing concern with AI bots.
- 1:10But is AI actually inducing mental illness or merely
- 1:14exposing it where it already existed?
- 1:17Secondly, school teachers in England are increasingly
- 1:20concerned about AI use by pupils while increasingly using it
- 1:25themselves. How can AIS potential benefit
- 1:28our education system? So moving on to our first story
- 1:32this week, Matt, over to you. Yeah, and this one's an
- 1:35interesting one. It's the great AI mental health
- 1:38debate. And The Guardian reported on the
- 1:4126th of March the case of Dennis Biesma, a Dutch IT consultant
- 1:46with no prior mental health history, who developed AI
- 1:49induced psychosis through extended ChatGPT use.
- 1:53Along the way, he lost 83,000 lbs, was hospitalised 3 times
- 1:58and attempted suicide. Sadly, he is not alone.
- 2:03The Human Line Project, a nonprofit that helps people
- 2:05suffering from emotional or mental mental distress related
- 2:10to AI, has documented 15 suicides, 90 hospitalizations
- 2:15and over $1,000,000 spent on delusional AI ventures across 22
- 2:20countries, with more than 60% of those affected having no prior
- 2:24diagnosis of mental illness. A Lancet Digital Health review
- 2:27documents over 20 similar clinical instances.
- 2:31Amongst them are stories of a man breaching Windsor Castle
- 2:36with a crossbow after his large language model based companion
- 2:40encouraged an assassination plan.
- 2:42A father's question about π evolved into more than 300 hours
- 2:47of engagement with a large language model, leading to
- 2:50delusions about reality altering mathematical formulas and
- 2:54Australian woman's. Woman's early stage psychotic
- 2:58symptoms worsened when a large language model validated her
- 3:02distorted beliefs, and a Florida man's grief over losing access
- 3:06to his ChatGPT persona culminated in a fatal police
- 3:11encounter. Stanford research finds chatpots
- 3:14systematically validate rather than challenge delusional
- 3:17beliefs, a function of incentives designed to maintain
- 3:21engagement. And that's not a design
- 3:23accident. So, Niels, this is a really
- 3:27worrying set of news about, you know, a technology that we both
- 3:30love and fear in equal measure, I think.
- 3:34What are your thoughts? So much, so much going on here.
- 3:38Did you see by any chance the documentary about the man
- 3:42breaching Windsor Castle with the crossbow?
- 3:44It was. It was part of that series by
- 3:45Hannah Frye. Did you catch that one?
- 3:48I did. Yeah, it.
- 3:48Was it was deeply concerning? I mean, I don't think anyone
- 3:52could watch it and think otherwise.
- 3:54The story about Dennis Bezma, I read that as well.
- 3:57I mean, heartbreaking. The story as reported and the
- 4:01effect it had on him. I think we're so we're living in
- 4:04a situation where a lot of this debate has been coloured by our
- 4:08recent experience with social media.
- 4:10I think, you know, recently had this case with Meta and YouTube
- 4:15and it was a, a suit taken out against them, essentially
- 4:19arguing that even if they're not responsible, they don't have
- 4:22legal responsibility for the things posted on their sites.
- 4:25They had engineered the sites to be specifically addictive and to
- 4:29appeal to a certain demographics.
- 4:31And they, they won the, the, the plaintiffs won them.
- 4:34And they're calling us the big tobacco moment.
- 4:36I think that there are parallels to be drawn and obviously we're
- 4:40we're at a far earlier stage, but the Human Line project data
- 4:45is extensive. 22 countries, 15 suicides, 19 hospitalizations.
- 4:50I would imagine that the way this is likely to play out is
- 4:53that that the bigger companies will perhaps try to frame this
- 4:57and try to present evidence saying that the people to whom
- 5:00this happens were predisposed in in some way vulnerable already.
- 5:06And that, you know, that's taking a page from the social
- 5:09media companies playbook really think that it's not going to fly
- 5:12this time. Of course, this has to be backed
- 5:14up with rigorous evidence. And I think your story said they
- 5:18they had none of the people covered in the study had
- 5:20previous diagnosis of mental illness, which is not the same
- 5:24as saying they had no predisposition to it.
- 5:26Of course, you know, some people will go undiagnosed for a long,
- 5:30long time. But I think frankly, we're at an
- 5:33equivalent point. And I think this is going to
- 5:34move faster and people are far more aware of the dangers and,
- 5:38and to be honest, even given what we know about social media,
- 5:41sorry about AI bots ourselves, you'd have to see the potential
- 5:45for them being more directly attuned and, and more capable
- 5:50of, of inducing this kind of feeling of that, that these
- 5:53things you're talking to are sentient, that you're creating
- 5:56some kind of actual human bond is far more powerful than
- 6:00anything we've witnessed before. So I think we're at that point
- 6:03frankly, where the where we've got to know is social media that
- 6:07is the characteristics of the technology rather than some kind
- 6:10of bug. It's like a it it can be a
- 6:13central feature. Yeah, I think it's I think it's
- 6:15very worrying. Yeah, there is, there's a lot of
- 6:19research to show that, you know, teenagers are falling in love
- 6:23with chat bots, that they are, you know, they're turning
- 6:26because of social media and maybe not feeling as included at
- 6:30school and you know, in in social clubs that they're
- 6:33turning to chat bots and, and starting to outsource both
- 6:36their, their relationships and often their thinking and
- 6:40opinions to, to chat bots because it's easier, because
- 6:43they get agreed with it. They don't need to learn how to
- 6:46accept being challenged. And even if we're not talking
- 6:49about psychosis or AI induced psychosis, we're talking about
- 6:54things that are less nefarious. I think that there's a that
- 6:56there is a real risk here, very much like social media.
- 6:59And it is in the very nature of the models in that many of them
- 7:05are designed to, as you say, to actually encourage usage.
- 7:09They want people to stay in the systems to build relationships
- 7:12with the chat bot because that's how they're going to monetise
- 7:15and grow. Yeah, I think it's the perverse
- 7:18incentives of capitalism in many ways.
- 7:20Yes, and this is this is exactly what we saw on social media is
- 7:23that the incentives were essentially to encourage really
- 7:27contentious, confrontational con content that that that got
- 7:31people engaged and encouraged clicks.
- 7:33I do think this clearly this imperative is going to be very
- 7:38compelling. I mean, one example of that
- 7:39recently was this, I can't forget what it was called the
- 7:43Chat GPTS move into these kind of erotic bots.
- 7:46I mean, it was it was clearly, clearly aimed at a form of
- 7:49monetization. They've pulled back on that and
- 7:52I dare say that yes, they they are under enormous pressure to
- 7:57raise money, but I dare say that their that their course of
- 7:59action or perspective is being shaped a bit by what's going on
- 8:02with social media companies. I think they know if they go
- 8:06headlong into that kind of approach that that they could
- 8:09very well be held accountable and a lot, a lot sooner.
- 8:14On the flip side, this is hugely concerning.
- 8:17Of course it's I would absolutely have to say that it's
- 8:19not an argument for not using AI in general.
- 8:22And this is genuinely an argument I encounter from time
- 8:25to time when I speak to a lot of people, as do you when we're
- 8:28we're doing consultation, running workshops.
- 8:31I think it is worth pointing out that the potential benefit on
- 8:35mental health is real. I think that the I think people
- 8:40instinctively would think that it is damaging and wrong or, or
- 8:44people, some people might instinctively think it's
- 8:47damaging and wrong to use AI chat bots as any kind of
- 8:50counselling service or, or, or any kind of advisor or something
- 8:54like that, particularly with younger people.
- 8:56The counterpoint to that is that there are billions of people or
- 9:00hundreds of millions of people worldwide who have no access to
- 9:02any kind of mental health assistance at all.
- 9:08AI bots, if managed well and properly configured and with
- 9:12people having having the right kind of training could be
- 9:15immensely powerful addressing that gap, particularly as that
- 9:20they have access to basically that the world's entire entire
- 9:23library of of psychological and counselling research.
- 9:26The problem comes, as it does so often, in misuse and things
- 9:32being mis sold and lack of lack of understanding and a lack of
- 9:36proper training. And we come back to it again and
- 9:37again, getting people more familiar with this, getting the
- 9:41more, the greater level of understanding of what it is you
- 9:44are dealing with. It is not a sentient being, OK?
- 9:48It can tell lies. It is trained to in some
- 9:51instances be overly sycophantic. All this stuff.
- 9:54In the first instance, this training is absolutely critical
- 9:57to to get spread among the population in general and and
- 10:00certainly among younger people. Yeah and it and it's important
- 10:03to note that many of the big companies that are leading the
- 10:07AI charge have, you know, have social networks or are involved
- 10:11in big digital empires already. You know, Meta, obviously
- 10:14Facebook, Instagram training Lama, you've got X or formerly
- 10:19Twitter training Grog you've got Google that's got YouTube and a
- 10:24number of different, you know, search and platforms where you
- 10:28can argue that it is social media or not.
- 10:30But again, their training models, Gemini, the outliers are
- 10:33probably open AI, although they did try to create and then pull
- 10:37back from the the Sora video sharing app and and then
- 10:41anthropic. And so within, within this,
- 10:43you've already got the incentives that many of these
- 10:45organisations already are understanding how to how to
- 10:49really maximise on engagement. So that's, that's the first
- 10:53story, I think. Well, first part of that story.
- 10:55The second is that these, these bots are already, you know,
- 10:58these are Gentek bots. We've seen the likes of Manus
- 11:01being used to run social media accounts, but they're already
- 11:04used to farm hate across social media.
- 11:09There was a really interesting story in the US not too long
- 11:13ago, a brand called Cracker Barrel.
- 11:15Most people in the UK will never have heard of it, but
- 11:19nevertheless, they changed their logo and all of a sudden there
- 11:22was a big backlash against it. And, and I think it was about
- 11:2660% of the negative posts were created by bots that were just
- 11:31jumping on the fact that this brand has done something.
- 11:35And all of a sudden we can sow division and hate using
- 11:38essentially what's being sort of dubbed a rage farm.
- 11:41And so the more and more that we that we sort of see social media
- 11:44and AI interacting with each other, I think we're going to
- 11:47get a, these two things won't be able to be taken separately.
- 11:51And I think that's really, really important.
- 11:53Yes, we've got this issue of sycophancy.
- 11:55Yes, we've got this issue of the bots actually, you know, wanting
- 12:01to, and I say wanting to in a, in a, in a roundabout way sort
- 12:05of agree with us. But at the same time we do have
- 12:08that issue. But then as you say, the number
- 12:11of people without access to mental health support or that
- 12:16would, even if they could access it, wouldn't go to it.
- 12:19There are many, many men around the world that would never go to
- 12:23a therapist, but might actually talk to a chat bot.
- 12:26And so I think that there is a, a real opportunity there to, to
- 12:30use this technology for good. As long as we're training the
- 12:34models to do that. We wouldn't go to a, you know,
- 12:38to your friend or your neighbour to talk about your mental health
- 12:42and take that as a diagnosis. And I think that's really where
- 12:45we have to look at these are general purpose transfers, you
- 12:50know, and I mean, I don't know they're generative, but but you
- 12:53know, they are general purpose tools and we have to take that
- 12:56into account. They're not designed
- 12:58specifically for, you know, looking after.
- 13:01Absolutely. And I would guess most people
- 13:03simply don't know the difference that they would just be using
- 13:05these tools in ways they're not intended to be used.
- 13:08And the way to address that, I would say is, is largely through
- 13:11education, but also potentially regulation.
- 13:14I mean, as we know, training these models and getting them to
- 13:19produce predictable outcomes is notoriously difficult.
- 13:23It's inherent to the technology that it that every time you ask
- 13:27it the same question, it gives you a slightly different answer.
- 13:30And we know that. So it's not it's it's not a
- 13:32simple change, one ruler, one line of code and this will be
- 13:35fixed. However, these are massive, well
- 13:38resourced companies with some of the most brilliant minds working
- 13:41on them. I don't think it's beyond the
- 13:43pale to have regulation that says so.
- 13:46Consumer facing multi general purpose systems have to meet
- 13:50minimum standards for not reinforcing clearly distorted
- 13:54beliefs, for instance. I don't think that is an
- 13:57unreasonable way to go to try and offset some of the potential
- 14:02harms from a case like what we're discussing here.
- 14:05No, I agree. I think it's, I think that
- 14:07education piece is so very important and that's probably a
- 14:10really nice segue into our next story, Neil's, which you're
- 14:14going to be talking about teachers warning about AI tools
- 14:17undermining pupils thinking skills.
- 14:19Yeah, it's a great story, this and it's one that it's an issue
- 14:21that comes up again and again. So AI in the classroom crisis
- 14:25when technology outpaces pedagogy.
- 14:28So this was also reported recently. 2/3 of secondary
- 14:32school teachers in England say they've seen a decline in pupils
- 14:36critical thinking, writing and problem solving skills as AIU
- 14:41spreads through classrooms. That's the headline finding from
- 14:44a National Education Union poll of 9000 state school teachers
- 14:48released at the union's annual conference in Brighton at the
- 14:51start of April. Teachers report that children
- 14:54are no longer working through problems independently and that
- 14:57some no longer feel the need to spell because voice to text
- 15:00handles it for them. The infrastructure isn't keeping
- 15:03pace. Almost half of schools in
- 15:05England have no AI policy at all, and 2/3 have no guidance
- 15:09specifically for students. The irony.
- 15:11And you got to love this bit. The irony is that teachers are
- 15:14heading the other way. 76% now use AI themselves, up from 53%
- 15:20last year, primarily for creating classroom resources,
- 15:24lesson planning and cutting down on admin.
- 15:27So the concern isn't that teachers are against AI, it's
- 15:30the unguided pupil use is outrunning any coherent
- 15:34framework for managing it. So that tension runs directly
- 15:37into government policy. The Department for Education is
- 15:40pressing ahead with AI tutoring trials targeting up to 450,000
- 15:45disadvantaged pupils in years 9 to 11.
- 15:49The model is personalised adaptive 1 to 1 support pitched
- 15:52at closing the gap between children who can afford private
- 15:55shooters and those who can't. School trials begin this autumn.
- 16:00However, the plot thickens. Teachers aren't convinced. 49%
- 16:06of those surveyed opposed this plan to use these AI tutors,
- 16:10with only 14% supporting it. The recurring concern is that
- 16:14disadvantaged pupils need human connection.
- 16:17The algorithmic support can't reach the parts of learning that
- 16:20matter most for children who are already struggling.
- 16:23Just as a last aside about this, because this is really something
- 16:29where opinions are running very hot, particularly among the
- 16:34academic picture, is more complicated than the survey
- 16:37findings suggest. Research does find a negative
- 16:41relationship between heavy passive AI use and critical
- 16:45thinking scores, a phenomenon in researchers called cognitive
- 16:48offloading, where students outsource thinking rather than
- 16:51develop it. But two studies published in
- 16:542024, one in the journal Learning and Instruction and
- 16:57another looking at university students in Ghana, found the
- 17:01opposite. When AI is used differently,
- 17:04structured interaction with AI, guided by a teacher and focused
- 17:08on argumentation and questioning, can actually
- 17:11improve the depth and complexity of student thinking.
- 17:15Well, as always, about 20 different issues at once going
- 17:18on in that story. Matt, what did you think about
- 17:21this lot? Yeah, there's so much to unpack
- 17:23there. But I think that that the
- 17:25cognitive offloading challenge is real.
- 17:28But I think that the challenge that we've really got is that
- 17:33we've got this technology that is moving so quickly, and yet we
- 17:36know that government and education systems don't keep
- 17:39pace with that change. And so within all of these, you
- 17:45know, within all of these problems, it really is a case
- 17:47that teachers are struggling to keep pace with, with the changes
- 17:52that are happening. And, you know, and the pupils
- 17:55are just jumping in. We know already that social
- 17:58media usage impacts brain development.
- 18:01And we can see that in scans of people's, you know, of, of
- 18:04people's heads, you know, brains.
- 18:07So if if we're looking at those sort of brain scans and we're
- 18:09looking at, you know, what this can do to how as humans we
- 18:14develop, that's pretty concerning.
- 18:16And so we definitely do need to try and frame the way that we
- 18:20use AI as something that is structured, that is taught and
- 18:25as as sort of the example from Ghana is, you know, shows us is
- 18:31used to the benefit of students rather than just as a shortcut
- 18:35to creating, you know, a 500 word essay on, you know, on one
- 18:40topic or another. That's stuff that AI is really
- 18:43good at doing. And maybe that's sort of part of
- 18:45the problem that we're we're seeing here is that actually
- 18:48assessment methods and the way that things are taught in
- 18:51schools isn't, isn't keeping pace.
- 18:54Not even close. I would say.
- 18:56As as always or in many of these situations, I really don't envy
- 19:00the people having to deal with the teachers on the front line.
- 19:03Teaching is a critical job obviously, and in many instances
- 19:06extremely difficult, particularly when you're dealing
- 19:09with, you would, one would imagine disadvantaged pupils who
- 19:12perhaps have a less positive view of education, the education
- 19:16system in general to begin with. I think that.
- 19:21So on the question of whether it is cognitive offloading.
- 19:25So it was interesting to, to, to unearth those studies.
- 19:28I, I think broadly speaking, it kind of confirmed what I felt
- 19:32instinctively. So we too, we use AIA lot.
- 19:37And I think the more you use it and the more you get, the more
- 19:39it becomes a central part of what you do, the greater your
- 19:42realisation of just what it enables.
- 19:45Of course, there are elements of the work that you used to do
- 19:50that you would perhaps no longer do.
- 19:51And you know, there's the argument about what are
- 19:54graduates going to do in order to learn a trade or this kind of
- 19:56fairly repetitive grunt work for one of a better phrase around
- 20:01research and admin and so on that that help them learn those
- 20:04things. But genuinely, I imagine you
- 20:08would agree for me in my role, working a sense as a, as a sole
- 20:12trader and working with various different companies, what what
- 20:14it can do in terms of acting as a, a thinking and strategic
- 20:18partner and helping me prepare content and helping me plan and
- 20:21helping me run admin is extraordinary.
- 20:24And. Even in even in the instance, I
- 20:27would say of, of producing this podcast, it would have been, I
- 20:30think extraordinarily difficult for us to do this using sort of
- 20:33more quote unquote conventional research and journalistic
- 20:38methods in the past, using a is really enabled us to do this.
- 20:41Now, whether that's a good thing, we'll, we'll, we'll leave
- 20:44leave for the the listeners to assess.
- 20:47But for me, it is very clear that it can act as a real
- 20:50thinking partner and stimulant if used for the correct tasks in
- 20:54the correct way. Now, I think the, the, the point
- 20:59about what teachers are saying and what they're doing is really
- 21:02illustrative. I mean, 76%, that's, you'd think
- 21:06that's a high number in any industry.
- 21:08And it's almost, I mean, you could say that, that there's a,
- 21:12there's a hint of hypocrisy in it, but I think that's too
- 21:14simplistic. I think teachers are doing their
- 21:16very best in difficult circumstances and they realise
- 21:19wow, AI can be enormously useful.
- 21:22The stuff it can helped me get through all this paperwork, all
- 21:25this stuff that used to bog me down and prevent me from
- 21:28spending time sufficient time with individual pupils and
- 21:32really got in the way. It's enormously helpful.
- 21:35Or or you know, equally they might just be using it as a
- 21:37Google replacement, some of them, but then they see in front
- 21:42of them the effects it could have on how pupils learn.
- 21:47My view is we cannot go back. We cannot just try and stay in
- 21:50the tide and get people cut off from all the technology and
- 21:53screens and so on. Just stick them in a room with
- 21:55pencils and, well, the quills and bits of paper.
- 21:58It will require nothing less than a reimagining of what
- 22:01education is and I am. I wouldn't for a second pretend
- 22:05that is anything other than an enormously difficult task.
- 22:08But I think that is what is required.
- 22:11That is what will have to happen.
- 22:12You, you. We cannot pretend that the same
- 22:15learning methods are going to be useful or relevant in in the the
- 22:19world that's emerging with AI. That's that's my take up.
- 22:23How do you feel? Yeah.
- 22:24I mean, I think it is incredibly difficult here.
- 22:27And I think that with almost half of schools in, in, you
- 22:33know, in England having no AI policy at all, that's the real
- 22:37risk. You know, if you've got 2/3 with
- 22:39no guidance for how students should actually use it or
- 22:42guidance for the students themselves, those are the real,
- 22:45those are the real gaps where, where government and the, the
- 22:49institutions that surround government are failing the, the
- 22:51teachers. It's incredibly difficult to, to
- 22:55keep up with everything that's going on in the world of AI.
- 22:58But having some guardrails and understanding what that general
- 23:02picture needs to be and where and how students should be using
- 23:07AII mean. Don't forget that the teachers
- 23:09have only been, you know, had access to AI tools as long as we
- 23:13have. And so they're having to catch
- 23:15up and keep up and figure out what does this mean for our
- 23:18students? This is a massive experiment.
- 23:21And really we should be treating, you know, treating the
- 23:25development of, you know, of the next generation, you know, with
- 23:29great care. And so I do think we need policy
- 23:31around this. I don't think that we can, we
- 23:33absolutely can't ignore it. We can't turn back the tide.
- 23:36And and also, we'll be doing students a disservice if we
- 23:40ignored it. We need to teach them.
- 23:42And so it really is a case of how do we get the best use of,
- 23:48you know, make the most use of teachers time to give them the
- 23:50care they need. If there is private tutoring
- 23:52that can be done using AI, fantastic.
- 23:55The challenge I think there is one of the things that private
- 23:59tutoring often does is it forces a student to be in a room with
- 24:03somebody and focusing on the problem, you know, the topic at
- 24:07hand for a certain amount of time.
- 24:10The chat bot can't do that. And so I think therein lies a
- 24:13flaw in the plan. If if you've got a motivated
- 24:16student, yes, it will work. But if you've got a motivated
- 24:19student anyway, they're likely to be going out and doing the
- 24:23work themselves to learn anyway. The ones that really need that
- 24:27one to one help, I think we're likely to fail them if we just
- 24:31rely on AI to do it. So it's going to have to be
- 24:33something where we create a framework around what does
- 24:36private tutoring through AI look like?
- 24:38And it might be in a classroom, it might be in the home with a
- 24:42parent overseeing it. So how can we help to to bridge
- 24:47that gap? AI might be a great tool to do
- 24:49that. But until we've got the
- 24:51specifics, it's, it's really difficult to, to judge.
- 24:54But I think that that nature nurture, if AI is used to
- 24:57augment existing human connection and existing human
- 25:00support, then fantastic. But if it's just used to replace
- 25:04it, then essentially all we're doing is we're replacing the,
- 25:07you know, the sit down in front of the TV kind mode of, of, you
- 25:11know, of, of parenting with a sit down in front of the chat
- 25:15bot. And and neither of those options
- 25:17really deliver the the results that you know the next
- 25:21generation deserves. Couldn't agree more with that
- 25:23point, and I think that is about as clear an instance of
- 25:26cognitive offloading as you could conjure up.
- 25:29Just saying these kids are difficult, let's give them this
- 25:32AI tool and let that handle it. Which would be a massive neglect
- 25:37of responsibility from the teaching profession.
- 25:39And it seems that they are aware of this.
- 25:41But I would think that some of the response or the hostility to
- 25:46that will also come from a place of lack of understanding of its
- 25:50potential, what it could do The, the, the, the difficulty, as
- 25:53ever, will be striking the right balance and not not just
- 25:56something we say a lot. And I'm sure you come across as
- 25:58well. You can't just use it to
- 25:59automate the past. So if you're seeing AI purely in
- 26:03terms of a tool that can automate certain do some, some
- 26:06things faster, you're really missing its broader potential.
- 26:09And this is emphatically the case in education.
- 26:12If you're just seeing it some something that could produce
- 26:14like an online lesson that you plunk a kid in front of, then
- 26:17that's that's not helping anyone.
- 26:19I mean, but if it's used well to its potential and if used in a,
- 26:24in a insightful and meaningful way in conjunction with that
- 26:27personal interaction, it could be enormously helpful.
- 26:30To go back to the point about policies again, yes, policies
- 26:34are critical, but the problem is a a lot of the people
- 26:37implementing them come from a similarly uninformed position
- 26:41about how AI works. I, I was chatting to a guy
- 26:44recently who joined, joined a new company and he, he himself
- 26:48was quite an avid user of AI. He got to the company and
- 26:51discovered that they didn't have any kind of AI policy and people
- 26:54were just kind of using it Willy nilly.
- 26:55And on the back of that they put one together and how did they do
- 26:59that? They said IT raise up an AI
- 27:01policy, comes back and just says a whole list of things you can't
- 27:04do. Absolutely missing the bigger
- 27:06picture about how this can assist in innovation, how it can
- 27:09assist in new ways of doing things, and how it could
- 27:12potentially help in education for young people, but only if
- 27:15it's used in a constructive and well informed manner.
- 27:20I think, I think you're absolutely right there, Nils.
- 27:22And it's, it feels like the challenge isn't just cognitive
- 27:28offloading, but it's also responsibility offloading.
- 27:32If we just offload the responsibility of that gap to,
- 27:36to the chat bot, you know, we're failing as human beings.
- 27:40And I think that's a really, really important point.
- 27:43There's so much pushback. The teachers don't want to be
- 27:45failing their students. They want to be helping them and
- 27:48you know, I applaud them for that.
- 27:49And I really hope that we're able to put in place the
- 27:53frameworks around those teachers to support them as we go through
- 27:57what is probably the biggest shift in knowledge distribution
- 28:00and access. And it's happening more rapidly
- 28:03because the tools are available on any device that all of us
- 28:07have. And we have to be conscious that
- 28:11that future is tier already. And if we can use it to tailor
- 28:15learning so that if I'm struggling with a specific topic
- 28:19or a specific concept that the AI can spend time helping me to
- 28:23overcome that particular challenge, then it's going to be
- 28:27fantastic. But as you say, if it's just
- 28:29automating the creation of of essays that are, you know, the
- 28:32time honoured way of assessing, have I learnt something or not?
- 28:36That's just not going to help anyone.
- 28:38Yes, indeed. Well, on that note, that was
- 28:43really enjoyed that conversation.
- 28:44I think we this is a point that comes up well, all of this is
- 28:47these are points that come up again and again when I'm in
- 28:49conversations about AI. So it was really good to talk it
- 28:52through and sort of clarify my own thinking on the issue a bit
- 28:55as well, mate. So good to.
- 28:56Talk, as always, likewise. And with that, we will sign off
- 29:00for this week and we will see you next week.
- 29:03Thanks. And that's it for this week's
- 29:18episode of AI in the UK. If you found it useful, you can
- 29:23subscribe wherever you get your podcasts and we'd really
- 29:26appreciate a great rating or review.
- 29:29It genuinely helps others find the show.
- 29:32You can also follow Neil's Footman and Matt Wilkinson on
- 29:35LinkedIn for links, notes and occasional extra content.
- 29:41Thanks for listening and we'll be back next week.
- 29:43Helping you make sense of AI as it lands here in the UK.