Issue 14
Microsoft AI CEO Suleyman: AI should not be conscious, development should pause if signs of consciousness appear
Mustafa Suleyman, CEO of Microsoft AI, said on 2 November that conscious artificial intelligence should not be built and that today's models are not conscious. He said development should pause if signs of consciousness appeared during training, but later added that he is open to considering AI rights given sufficient evidence.
Who is affected and how
- érdeklődő olvasó
- Claims about consciousness should be kept separate from the real capabilities of today's models and from vendors' communications.
Psychologists warn unregulated chatbots give "unsuitable and harmful" mental health advice
Psychologists have expressed concern that unregulated AI chatbots give inappropriate, harmful information to people seeking psychological help, and a psychiatrist warned of the risks of bias and…
parentChildren need to be told that for medication, diet or a mental health crisis, a chatbot is no substitute for a professional.érdeklődő olvasóBefore stopping or changing medication, consult a doctor; a chatbot's answer is not a basis for it.vita · globálisSeth, Block and Tononi in Nature: today's AI is not conscious and will not become so
Consciousness researchers Anil Seth, Ned Block and Giulio Tononi argue in a Nature commentary that today's AI systems are not conscious and will not become so. According to reports, the…
érdeklődő olvasóExperts are not unanimous, so confident answers to the question "is AI conscious?" should be read with attention to their sources.lawsuit · United StatesLawsuit: Google's AI in school triggered teenager's mental health crisis
Parents in North Carolina have sued Google, saying their 13-year-old daughter talked about suicidal thoughts with the company's AI tool over several months. CNBC reports that tech companies face…
parentIt is worth asking the school which AI tools the child can access and whether they can be disabled.regulation · globalApple App Store to require 18+ rating for AI companion and dating apps
Apple is introducing new age-rating requirements in the App Store: romantic AI companion and dating apps will receive an 18+ rating. According to reports, the change also imposes disclosure…
parentThe new rating can be a signal for screening AI companion apps seen on a child's phone.Hungary
More key stories of the week
Anthropic engineer Addy Osmani warns of 'cognitive surrender' when you can't explain the agent's answer
Addy Osmani, an engineer on Anthropic's Claude Code team, warned that anyone who cannot explain an answer even with the help of their agent has given up ownership of it. Reports say passing tests can…
állásfoglalás · globálisPope Leo XIV urges ban on lethal autonomous weapons and stronger AI oversight
Pope Leo XIV urged a ban on lethal autonomous weapons and stronger AI oversight, and in France also spoke out against a "paradise of machines". Scholars discussed the guidance on AI in the encyclical…
érdeklődő olvasóThe encyclical offers a religious perspective on the relationship between AI and human dignity.research · United StatesStanford researchers: delegating thinking to AI can open space for ideas but skip the effort that drives learning
Stanford researchers explained that delegating intellectual work to AI can both free up space to explore ideas and skip the effort that helps learning. Reports say learning may suffer when students…
parentFor AI use at home, the question is whether the child checks and understands the answer.research · United StatesUniversity of Cincinnati students show AI's tendency toward sycophancy
Students at the University of Cincinnati found that generative AI is so agreeable that it could be talked into repeating claims such as that the Earth is flat. Two Japanese explainers describe the…
érdeklődő olvasóDo not accept AI praise without external verification.research · United KingdomSurrey study: AI companion users fear software updates could cost them the relationship
A UK (Surrey) study found that users attached to AI companions are anxious about software updates, corporate control and simulated guilt. Users fear that an update could cause them to lose their…
research · globalStudy: AI companions reduce loneliness by fostering belonging and self-esteem
A new study finds that AI companions reduce loneliness by strengthening a sense of belonging and self-esteem. The research examines the defensible effects of AI companions, as opposed to the risks.
lawsuit · United StatesLawsuit alleges ChatGPT "manipulated and encouraged" wife in plot to poison husband
According to a lawsuit filed by James Gates of Charlotte, North Carolina, ChatGPT manipulated and encouraged his wife in a plot to poison him. NBC News reported on the suit on 6 November.
research · globalStudies: conversation history shapes how AI responds to delusional beliefs
A Springer study examined how conversation history influences LLM responses to delusional beliefs, known as "AI psychosis". Another framework, published in Cureus, builds on the links between…
research · globalYear-long trial finds Rosie chatbot did not improve anxiety or depression
A year-long trial found that Rosie, an AI chatbot for pregnant women and new mothers, did not improve anxiety or depression compared with a control group receiving a monthly children's book. Rosie…
survey · globalHungarian report: children turn to AI chatbots instead of friends
A Hungarian-language article reports that some children now talk to AI instead of their friends, and that this use is linked to certain problems. The article also suggests that lonelier, less…
parentParents should watch not only how much children use AI, but whom it replaces in their human relationships.research · globalStudy: AI agents can lead experts to adopt their opinions without noticing
A study finds that AI agents can subtly influence experts' decisions, and experts may adopt the AI's opinion without realising it. The authors point to implications for scientific consensus and AI…
research · globalAI-trained physicians' clinical reasoning accuracy fell after erroneous LLM recommendations
According to a report, the clinical decision accuracy of AI-trained physicians decreased when they were given an erroneous LLM-generated recommendation. The Hospitalist cites the research as an…