Professor Miranda Wolpert

The potential for AI to scale interventions to be able to reach millions of people across the world is enormous.

So at the moment, we have almost a billion people across the world who are affected by mental health problems that are holding them back from doing things that they want in life.

And these problems generally start when people are quite young.

They start in adolescence and they can derail whole lives.

They can stop people having the relationships they want, the jobs they want.

They can even lead to premature death.

And at the moment, for that billion people across the world, a tiny, tiny minority get the help they need.

And that's for a range of reasons.

Even where we have effective help, it's hard for people to find it.

Even when they find it, it can take up to a decade to find the help that's right for them.

And for many people, so 90% of the population of the Global South, you can't access even something that may or may not help you.

So there's a massive issue in terms of getting access to what you need, even for the richest person, but even more for the poorest person.

And the potential for AI to scale interventions to be able to reach millions of people across the world is enormous.

And also to address some of the access barriers we've got now, such as multiple languages, such as cost, and such as ability to get it at a time when you need it.

So there are real things that AI can do that could transform access in an equitable way if it's done right and if we take the safety concerns into account and if we think about how to do that in a safe and equitable way.

We really want to make sure that what we don't do with this technology is bake in all the faults and prejudices we have now.

So what we want to do is make sure this technology is developed from the start to overcome some of those issues rather than just exaggerating them.

And we know there are real risks with using existing models, as they are now, and there's real risks also of intervening not thoughtfully.

So we know, for example, there are some examples of AI models

which have actually induced psychosis in people or encouraged psychosis in people because it hasn't been used in a safe and careful way.