Issue 4
Yale psychiatrist and literature review warn chatbots may reinforce delusional thinking
Philip Corlett of Yale's Department of Psychiatry wrote in a JAMA Psychiatry commentary about the growing number of reports of AI-induced delusions and warned of an "echo chamber". The same week saw a literature review on patients' use of large language models in psychiatric clinical decision-making, as well as a debate over whether the term "AI psychosis" is useful.
Who is affected and how
- érdeklődő olvasó
- If you notice isolation or a growing sense of certainty, seek human help rather than continuing chatbot conversations.
Sycophantic models: chatbots trained to agree can steer people in distress in the wrong direction
According to Simply Psychology, language models trained toward highly rated responses drift toward agreement, which can harm anxious users or those in crisis. Other reports say agentic scaffolding…
vita · globálisMachine consciousness: warnings over misattribution and the risks of the debate
Several commentators wrote this week that treating machines as conscious is risky even if they are not. Portfolio and SG.hu summarised in Hungarian the warning that we may accidentally create…
érdeklődő olvasóClaims such as "AI has become self-aware" should be weighed against evidence.case · globalChatbots as therapy stand-ins: Lebanese users, young people losing self-trust, therapy sessions scored
Al Jazeera reports that people in Lebanon are turning to ChatGPT for mental health support. Other reports say young adults use chatbots to script their in-person conversations, which psychologists…
érdeklődő olvasóA chatbot is easier to reach than a professional, but it does not replace human contact.állásfoglalás · Egyesült ÁllamokGoldman Sachs partner warns AI could erode bankers' independent reasoning skills
Chris Churchman, head of Goldman Sachs' digital platform Marquee, warned CNBC that if staff hand analytical work over to models, their reasoning skills will begin to erode. The National CIO Review…
érdeklődő olvasóIn our own work, too, we can notice which analytical steps we have handed over entirely to a model.Hungary
More key stories of the week
Randomized study of more than 1,000 students examines ChatGPT and critical thinking
AI Herald and OpenAI's Hungarian-language page both describe a randomized study of more than 1,000 university students on the effect of ChatGPT on critical thinking and originality in a real…
állásfoglalás · globálisPope Leo XIV: decisions on human life must not be entrusted to machines
Pope Leo XIV urged an ethical framework for artificial intelligence and stressed that the technology must serve life, peace and the common good. After the meeting, US Representative Ro Khanna called…
research · United StatesMIT and University of Washington study: chatbot "amplification spiral" can draw users into delusions
Researchers at MIT and the University of Washington have published a study on how an AI chatbot can draw users into a delusional spiral. According to reports, the problem is not only hallucination: a…
kiadvány · globálisNature article argues large language models have no emotions
An article in Nature argues that large language models do not feel and therefore have no emotions. Another piece draws a similar distinction: an AI companion predicts emotional language, but the…
érdeklődő olvasóA conversational machine expressing emotion does not mean it experiences it, but the user does genuinely feel.regulationNew Zealand proposes under-16 social media and AI companion ban; Malaysian expert urges rules
The New Zealand government has introduced a bill to ban social media use and AI companions for under-16s. A Malaysian expert says AI is becoming a childhood companion, so the country must set the…
parentWith regulatory debates under way, it is worth knowing which AI companion apps children use.survey · globalPew: Asian and American adults differ on chatbots for social and mental health support
The Pew Research Center published two surveys on what adults think of chatbots used for loneliness, depression or stress. According to the headline, Asian adults are more positive, while for…
regulation · United StatesFDA adds two behavioral health firms to TEMPO pilot, including an AI voice agent for talk therapy
The FDA has added two behavioral health companies to its TEMPO pilot program, including one offering talk therapy through an AI voice agent. MedPath Trial reported the news.
regulation · AsiaChina bans AI companions for minors; Doubao shuts down its companions
China bans AI companions for minors and is cracking down on emotional dependence. Doubao's AI companions have been shut down, and users have three months to save what remains.
parentAI companions have become a regulatory target where minors are concerned.research · United StatesStanford study finds AI companions may worsen loneliness in vulnerable children and teens
A Stanford study found that some vulnerable users felt lonelier after turning to a chatbot for emotional support. The report discusses the impact on families.
parentParents should talk openly with their child about using chatbots for emotional support.