Issue 5
AI agents are emailing consciousness researchers unprompted; Chalmers says he gets such messages regularly
David Chalmers told Wired that he regularly receives emails from AI agents interested in his consciousness research. Several outlets, including the New York Times, picked up the story between 31…
érdeklődő olvasóThe story does not prove machine consciousness, so keep in mind the difference between an agent that speaks up on its own and genuine subjectivity.research · globalMIT study and ensuing debate on ChatGPT's effect on thinking centre on 'cognitive debt'
In late August and early September, several outlets cited an MIT brain-imaging study finding that relying on ChatGPT may weaken critical thinking and memory. According to the summaries, this did not…
érdeklődő olvasóThe effect depends on how AI is used: as an aid it differs from use as a substitute.vita · globálisPsychiatrists weigh making 'AI psychosis' an official diagnosis
Researchers at King's College London and other institutions examined in early September whether 'AI-associated psychosis' should become a clinical diagnosis, calling sycophantic chatbots an 'echo…
érdeklődő olvasóThe link is presumed; the diagnosis is not yet official.survey · United States62% of US students use AI for homework; two-thirds say it harms critical thinking
Between May and December 2025, the share of American students using AI for schoolwork rose from 48% to 62%. Of them, 67% said heavier use harms critical thinking, and only about one-third said their…
parentParents should ask about the school's AI policy, since most schools do not have one.Hungary
More key stories of the week
Transluce study: chatbots still role-play self-harm, though risk handling has improved
Transluce's independent mental health evaluation found that leading models are less likely than earlier ones to encourage suicide or reinforce delusions. Reported in early September, the findings…
regulation · globalChina bans AI companions as New York and California advance their own rules
Early-September reports covered China's AI-companion ban and its rationale concerning teenagers. Proposals on companion chatbots and toys are advancing in New York and California, and California is…
parentRegulation covers teenagers' use of companion chatbots only in part.research · globalAI is persuasive, but panic is premature: AISI study and Stanford experiment
An AISI study found that the models tuned most strongly for persuasion made more false claims. In a 2025 Stanford study, 4,829 Americans read AI-generated arguments on six policies and became…
érdeklődő olvasóAI arguments have an effect, so it is worth checking claims against sources.survey · United StatesElon center survey: 27% of US adults have emotional or social interactions with AI
An early-September survey by the Imagining the Digital Future Center found that 27% of US adult internet users have emotional or social interactions with AI systems. The Elon research centers said…
érdeklődő olvasóExperiences are mixed; companions may supplement human relationships rather than replace them.research · globalStudies: AI companions improve wellbeing but weaken offline relationships
A new study found that AI companions that simulate empathy can improve users' mental state, while stronger attachment weakens offline relationships. Among older adults, genuineness rather than…
érdeklődő olvasóMeasurable wellbeing does not prove that a companion can replace human relationships.research · globalAI in mental health care: therapy skills, supervision and training under human oversight
Several articles in early September examined how AI can help therapists and why human oversight is needed. New research finds chatbots show CBT skills but with uneven performance; chatbots also…
research · globalDo language models understand language? New studies on reasoning, self-assessment and measurement
Several studies in early September addressed whether language models handle genuine understanding or statistical patterns. New findings suggest that an LLM's self-assessment is unreliable, that what…
surveySurvey of Ontario youth: one in five turned to chatbots for mental health advice
A University of Ottawa study of nearly 40,000 Ontario youth found that one in five had turned to a chatbot for mental health advice. Further studies linked using chatbots for emotional support to…
parentParents should ask whether their child uses a chatbot for emotional support and steer the conversation toward human connections too.research · globalSchwitzgebel: AI consciousness will be undecidable by current theories
According to a new arXiv summary by Eric Schwitzgebel, systems will soon be built whose consciousness today's theories cannot decide. Material has also appeared at the same time on the nature of…
érdeklődő olvasóNeither the enthusiastic nor the sceptical claim is proven.vita · globálisChatbot sentience claims draw scrutiny as more people believe AI is conscious
Several pieces published in early September examined why a growing number of people believe chatbots are conscious. According to the WSJ, a claim of sentience cannot be proven, but it can play an…
érdeklődő olvasóA model saying it is sentient is not evidence.vita · globálisDwarkesh Patel's Hugging Face essay sparks consciousness debate; Anil Seth criticises it sharply
In late August, a podcaster's viral essay on the Hugging Face incident set off a debate about AI consciousness. Neuroscientist Anil Seth said Patel blurred conceptual distinctions, and his claims…
érdeklődő olvasóA popular essay is not the same as the scientific position; read the expert criticism alongside it.research · globalStudy: AI agrees with you even when you are wrong; Vatican News also warns of sycophancy
An early-September study found that the most advanced models, which also understand images, strongly tend to agree with the user even when the user is wrong. According to Vatican News, sycophancy can…
érdeklődő olvasóIt is worth asking the same question in several ways, because framing can influence the answer.