Issue 12
Seth, Tononi and Block in Nature: today's AI is very unlikely to be conscious
In a joint opinion piece in Nature, Anil Seth, Giulio Tononi and Ned Block give three reasons why today's AI systems are not conscious and machine consciousness is very unlikely. David Chalmers says the arguments are not strong enough, and Seth replied that they are not claiming proof but criticising the overestimation of the probability of AI consciousness.
Who is affected and how
- érdeklődő olvasó
- According to three leading researchers, consciousness is not the same as intelligence, so convincing linguistic behaviour is not in itself evidence.
Hinton says AI may already be conscious and humanity is 'toast' unless systems care about us
Nobel-winning computer scientist Geoffrey Hinton said at the Starmus festival and in interviews that AI may already show signs of consciousness, and that humanity is "toast" unless it can make such…
érdeklődő olvasóThe claim comes from one of the best-known AI researchers but is strongly disputed scientifically, so consciousness and risk are worth treating separately.research · United StatesGeorgia Tech researchers find a way to reduce chatbots' reinforcement of users' delusions
Georgia Tech researchers have found a method to make AI chatbots less likely to reinforce users' delusions. The Research News Center reported the result on 20 October 2026.
regulation · United StatesAnthropic's AI-rights and model-welfare moves: Suleyman calls it a safety risk
AI rights have entered Anthropic's usage policy, while Microsoft's Mustafa Suleyman calls "model welfare" a safety risk. Experts say belief in chatbots' feelings could create a vulnerability to…
research · United StatesStudy: 'AI psychosis' is rare but real, and first-episode patients may be most vulnerable
A review of electronic health records at a large medical center suggests that use of conversational AI chatbots may be associated with worsening symptoms. The study says the condition is rare but…
parentFor a young person in a mental health crisis, checking chatbot use is a relevant question.Hungary
More key stories of the week
Harari says AI could make humans 'the cows of the future'; National Review disputes the view
Yuval Noah Harari says AI's autonomous decisions could undermine social trust and leave people intellectually comparable to cows. National Review argues that this doomerism underestimates people's…
érdeklődő olvasóThe debate is about whether delegating decisions to AI leads to a loss of human agency.állásfoglalás · globálisAI chatbots in mental health: they can help, but may also reinforce harmful thinking
Curtin University and a psychiatrist writing in The Independent said AI chatbots can help with mental health but may also reinforce harmful thinking patterns. The psychiatrist listed specific risks…
érdeklődő olvasóA chatbot can be supplementary support, but experts do not consider it a substitute for a professional.research · globalCognitive offloading and AI dependency: new reviews on learning and thinking
A Cureus scoping review, a study of students' perceptions and a Nature audio long read all address how AI substitutes for cognitive tasks. The pieces cover risks to critical thinking and learning…
érdeklődő olvasóResearch suggests that AI's effect on thinking depends on how it is used.research · globalSycophancy benchmarks show wide gaps between AI models, and safety degrades across languages
A multilingual study finds that sycophancy is linked to weaker safety behavior across languages, topics and models. Artificial Analysis's AA-PersonalityBench ranked 38 models on 30 traits, with a…
kiadvány · globálisPilots' experience may help avoid AI-driven deskilling
An article says pilots have long faced the risk of skill loss from relying on autopilot. Their experience may offer lessons for workers who use AI.
érdeklődő olvasóIn AI-reliant work, too, the skills that the machine takes over need to be deliberately maintained.research · United StatesTexas A&M study: training on viral junk posts dulls AI models' reasoning
A Texas A&M study finds that AI models trained on viral, low-quality social media posts show "brain rot". According to Digital Trends, the models skip reasoning steps and pick up worse ethics.
research · globalMonitorability of reasoning language models: Nature Machine Intelligence paper and Columbia preprint
A Nature Machine Intelligence paper examines how understandable and monitorable the reasoning tokens of step-by-step reasoning language models are. A Columbia preprint reports that flawed rewards can…
research · United StatesAI companions and loneliness: studies suggest they can reduce it, but press coverage focuses on risks
A working paper by Julian De Freitas and colleagues finds that AI companions reduced loneliness in several studies. One overview notes, however, that press coverage of chatbot companions typically…
research · United StatesChatbots and manipulated photos can plant false memories, MIT Media Lab research says
Citing MIT Media Lab research, press reports say AI chatbots and fake photos can create false memories in people. According to the reports, over the longer term this may also be linked to a decline…
érdeklődő olvasóConversations with chatbots and AI-generated images can also influence memory.regulation · AsiaChina regulates AI companions from 15 July 2026
China became the first country to regulate AI companions specifically, with the rule taking effect on 15 July 2026. Five government bodies are involved in the regulation, including the Cyberspace…
survey · United KingdomBarnardo's: 43% of UK children have turned to a chatbot for advice or emotional support
Barnardo's data show that 43% of UK children have already turned to an AI chatbot for advice or emotional support. A Hungarian-language site also carried the figure.
parentA significant share of children seek emotional support from chatbots, so it is worth talking openly about it.