You are watching The Context. It's time for our weekly segment, AI Decoded. Welcome to AI Decoded. That time of the week when we look in-depth. It's one of the most eye-catching stories from the world of artificial intelligence. First, Using AI, being in therapy, will show how NHS England are using AI chatbots to help their growing waiting lists for mental health support. And like here in The Guardian newspaper, we'll ask whether a chatbot is really equipped to tackle complex emotional needs. Next up, Forbes asks, Could AI-powered genital scans actually become a thing? A new startup has launched a sexual wellness tool for women, and it's encouraging users to submit photos of their partner's private parts. The company claim they patented AI wizardry and can analyze images of male genitalia for the presence of common sexually transmitted infections. The Daily Dot shows one man's solution for job interviews. A tech CEO posted a TikTok showing him using an AI clone during a mock job interview. We'll show you a bit of that in a moment. And in the Hollywood report, Mamma Mia musical star, Sarah Poyser said in a social media post that artificial intelligence has replaced her for a BBC production.

She shared a message from an unidentified production production company, which said they had the approval from the BBC to use an AI-generated voice, so they no longer needed her talent. Well, the BBC has responded by saying, In these very particular circumstances, and with the family's wishes in mind, we have agreed to use AI for a brief section to recreate a voice which can now no longer be heard. This will be clearly labeled within the film. Finally, in the Telegraph, could AI come up with a perfect pint? According to a recent study made, Artificial Intelligence, made a beer tastier than normal. Researchers trained a computer system to take average existing beers and elevate them to an elite drink after knowing exactly what chemicals to add. With me is Priya Lakhani, CEO of Century Tech, artificial intelligence company that develops AI-powered learning tools for schools. Great to see you. You're standing by. Ready for us first, though. Just want to show you this just before we begin and get your thoughts on it. It's basically a look at the potential use of AI in therapy. As waiting lists for NHS mental health support continue to grow. A new wave of startups building AI chatbot solutions.

They're hoping to help our reporter, Hannah Carpel. How's more?

How would you feel about opening up to a chatbot? Well, millions of people now do in what's known as AI therapy. Across Of England, the NHS is using the software to help tackle their growing waiting list for mental health support. One of the companies helping them is Limbic, who work with a third of NHS talking therapies in England.

We've taken this data with consent from the NHS, trained it to be able to identify with very high accuracy what your primary problems are. We were able to take your goal for therapy, what your primary issues that you wanted to really focus on during therapy are through this highly empathetic They take 24/7 always on chatbot.

But can we trust AI as therapy?

The first thing that we need to know is how secure the data is. But we also need to know, of course, that there is patient benefit. You need to know that this is better than doing something else, or better at least than doing nothing. Rather than having chatbot replace therapists, I think the idea of having them blended together is probably initially a way of reducing the number of sessions that you need.

Some patients are more comfortable talking to a chatbot.

Our AI was inherently non-judgmental. It did not have a race or agenda, so it could not clash with the race or gender of the individual seeking support.

As more of us talk to the AI, its database and understanding of communities will grow. So that means more reliance on artificial intelligence in our healthcare. Hannah Carpell, BBC News.

So interesting issue there. A similar story on the front page of The Guardian newspaper, on the inside pages. He checks in on me more than my friends and family. It's a quote there. Can AI therapists do better than the real thing? Priya, I'm going to ask you that question. What do you think of it?

Well, that's a really big question. I think that AI can serve as an addition to conventional therapy, and it can augment it. But what's really interesting is what we just saw there with the LIMBIK that the NHS is now deployed across many many parts of the NHS in terms of its Talking Therapies program, is very, very different to what's in the Guardian article and what The Guardian focuses on. It's really, really important to distinguish between this. Firstly, let's just give a bit of context to our audience. Demand for therapy is really, really high. So waiting list, you've got 20,000 people in the UK that are off work every month for mental ill health. A quarter of people, a quarter of patients seeking treatment wait more than 90 days, and about half of them deteriorate within that time. I know we're a global program, so I pulled out just a couple of stats just to tell people in the audience how severe this is. So worldwide, about 12 billion work days, 12 billion are lost every single year due to depression and anxiety. And the pandemic triggered a 25% increase in general anxiety and depression. This is obviously a really big problem.

It's caught governments unprepared. They only spend about 2% of their health budgets on mental health. How can AI help is the question. What we saw from the segment there was actually an app that has got medical device status when it comes to it's regulated. That device there, it's been proven to reduce admin time and clinical workload by about 50,000 clinical hours, the CEO has said. It's a support tool. It helps to diagnose mental health. Then the point is it augments the process. Then you've got a clinician, you've got an expert that you can then see. Got you. That's the crucial thing. That's there. The Guardian, very, very quickly, is about the tens of thousands of apps that are often unregulated. This is the difference. Chatbot Friends, where they've finetuned an app using GPT, for example, and people are finding solace in speaking to these apps, but there are all sorts of issues with them. Firstly, is your privacy maintained when you're talking to these apps? Actually, there were some sweeping reviews. Actually, the Mozzila Foundation found that they looked at 32 popular mental health apps and found 19 of them were failing to safeguard user privacy.

Then psychologists make this point that if we end up in a world, Lewis, where your best friend is actually on your app and you're going to this best friend AI to talk to them, even about anxiety and depression, which, by the way, there are so many benefits of that in many ways. People are very lonely who have a stigma when they think there's a stigma of seeking therapy might go to the app. The problem is that you might be exacerbating the fact that some people won't then relate to humans at all because humans are chaotic and a bit messy. The story goes into that, but they're very different things. It's not regulated, and I think that's what the audience has to be very clear about.

Brilliant.thank you very much.But.

Then you have those regulated apps.

Thank you very much for making that point clear. Very important. Next story. An AI app claiming to detect STIs from photos of genital tools is a privacy disaster from Forbes there.

Shall I tell you about this app? The problem with this app is that if you go to the website, it's called Kalmarra. It encourages someone who's about to get into an intimate situation. You're in that environment for the first time, maybe, and things are heating up. It's to whip out your smartphone. Glad you said smartphone. Whatever, and take a picture of male genitalia and to see if there's an STI. Now, I think the problem is there. Stis are a massive There's a massive problem. One million pregnant women estimated to be infected with syphilis in 2016. There were 350,000 adverse birth outcomes. There's all sorts of problems with HPV. This is clearly a problem that needs to be solved. Ai can help. They use essentially computer vision techniques It's an AI model called segmentation, where you look at, for example, skin, and you try and find in the pixels of a digital image an anomaly. If you're looking for a skin lesion. The AI technology can absolutely help, but the problem is that, firstly, people Are posting images into this app where there might be security and privacy concerns? I started reading it and thinking, actually, this is potentially a very good use of machine learning and AI.

But then I started reading about the company, and in the Forbes article, they claimed that the company itself said that it had some elite academic partnerships that was then removed from their website once Forbes started probing into it. The big universities have claimed that they don't have a partnership with the company. You can start to see now where apps are out there, and unless there's a lot of transparency about how they're working, where the data is going, is there independent validation of the data sets, and then the AI model being performance and all of these things, it cast doubt upon the AI app's claims. For the audience, the clear thing here is before submitting certain photographs or any photographs, I would say, into these apps, you need to...

It's a privacy, private parts privacy nightmare.

You know the research I did for that piece just there, I read the paper and saw all the images.Okay.

All right.That's what I'm going to say. I'm not going to ask you anymore about that. Let's carry on. Let's go to the Daily Dot. Actually, just before you talk about it, let's just take a look at this video. Hi, Aidan. How are you?

Where are you coming from today?

I'm doing great, thanks. And I'm in London. I've been looking forward to discussing this opportunity with you today.

Great stuff.

If it's okay with you, I'd like to dive straight into it. Can you tell me about a product that you've launched that you're proud of and why?

Certainly. One of the products I'm most proud of is a mobile app that we designed to stream online shopping for customers. It integrated AI-driven recommendations and a highly intuitive UI.

So there's a man on his laptop with an AI version of himself doing a mock interview.

That's going to be us. We're going to be replaced in the future. That's incredible. The app claims that 80% of people who use the app are hired. It's really interesting. Lots of people use ChatGPT to help write their resumes and cover letters. Some would say that's reasonably smart. About 11% are denied the job when the employer finds out they're using it. I think in some ways, Lewis, it can level the playing field for people who may struggle with writing applications. The cloning of the interview, I've got a real problem with that because you can misrepresent... The AI clone could misrepresent your capabilities, your personality, your behavioral skill set. Lots of companies have cultural interviews, we do, to ensure that culturally you fit, whereas if it's a clone, you're not going to get that from the person. You can see all sorts of problems with that. But I thought that was great that you had the video there because that's where this is potentially going. Amazing.

A familiar theme coming up next, AI Replaces Mamma Mia, musical star for a BBC production. What's going on here?

I'm really glad, actually, you added a piece to what was in this story where the BBC have have said in a statement that they're actually trying to replace a voice actually that no longer exists. I think that context is really important. But in the story, it's really talking about the significant threat to employment in this sector, for example, voiceovers. Now, SAG After had a bit of an update a few days ago. They negotiated and ratified actually the agreement when it comes to voiceover actors. With voiceover actors, they've included, for example, if you're going to say that it's a voice actor, it has to be a human being. People will be remunerated if they're the voiceover and it's translated into a different language. The article points to, actually in the UK, maybe we should be copying our friends overseas and there ought to be some union agreement potentially for artists to protect them. Because Clearly, it lowers cost of production. There's less recording required. There are going to be production companies in the future that aren't in this context using AI to replace a voice that no longer exists, but it's just going to do it because it's simply cheaper.

Fascinating. Right. Something a bit lighter to finish. Bier, this is in the telegraph. Bier made by artificial intelligence tastes better than a normal pint, study suggests.

Yeah, so they've taken over five years, expert taste panels. Imagine the data now. This is the important thing. Expert Taste Panel. So data, 50 criteria by 15 experts tasting about 250 beers. Then you've got 180,000 reviews. You can imagine all these data sets. What they've done is then they've looked at lab analysis of the chemical composition of the beers. The AI essentially just suggests alterations or amendments to make the beer tastier. They've actually managed to do that. I'm wondering where our drink is now, producer. I was thinking we would do a live tasting. What's going on? So, Santo, to those who are going to enjoy this new beer. It's a new technology, and why not use it? Although some people will always just want that fully 100% human-crafted beer.

No, this is definitely AI I can get on board with. Definitely fan of that. Right, we are out of time. Brilliant stuff. Thank you so much for that. And that's it. We are out of time. We'll do this again same time next week.