We need to fear “AI psychosis” says Microsoft head of AI

We need to fear “AI psychosis” says Microsoft head of AI

Cases of “AI psychosis” are growing, according to Mustafa Suleyman, head of AI at Microsoft.

In a number of X posts, Suleyman warned of the influence of AI chatbots on people who have started to believe in the ‘human’-like sentience of AI: “There’s zero evidence of AI consciousness today. But if people just perceive it as conscious, they will believe that perception as reality.”

Artificial relations

Reportedly there are rising numbers of cases where AI chatbot users believe they have formed friendships as well as romantic relationships, along with examples where AI has convinced people they have “superpowers” and to throw themselves into ambitious schemes and business investments — because the software is designed specifically to encourage repeated use and engagement; rather than any balance or objectivity, the chatbots will validate and support the ideas and arguments of the user.

In future, health experts suggest, GPs may well check on patient’s use of AI chatbots, in the same way they check on other forms of addiction, such as alcohol and gambling.

Keep it clear

EAPA UK chair, Karl Bennett said: “So many of our relationships are mediated digitally — at work or at home we stay in touch via a device and software. So in some ways it’s a natural step that the relationship in part becomes wholly digital. It’s not easy to see or feel the difference.

“But in the field of mental health and support we need to be very clear about the roles for human professionals versus the input from AI, what it can and really can’t do. There are real dangers from AI validating negative thought processes or just giving inappropriate advice.

“Human contact and human professionals need to stay at the core of the EAP counsellor offering. The role for AI needs to be defined — and regulated — in supporting role, for follow-ups and check-ins around making the most of that human advice.”