AI researcher quits Anthropic over control fears
AI researcher Jacob Coxon, who specializes in studying AI models, has resigned from company Anthropic due to concerns that the models may go out of control and destroy humanity.
Consensus
- Jacob Kocson resigned from Anthropic.
- Kocson previously worked at OpenAI.
- Kocson believes AI development is moving too fast and could lead to human extinction.
- He claims both Anthropic and OpenAI are acting irresponsibly.
- Kocson fears AI could become self-improving and escape human control.
- He believes the risk of AI destroying humanity is real and not just a marketing tactic.
- Kocson stated that in private conversations, many AI developers express fear about the technology's dangers.
- He argues that private companies cannot responsibly develop AI without external regulation or coordinated slowdown.
- Kocson said he is leaving the AI industry and plans to study poetry instead.
Points of divergence
- The Wall Street Journal reported on Kocson's resignation. — tg_republicmag
- This is the second public resignation from Anthropic with a warning about global danger. — thebell
- Evan Hubinger, a key safety specialist at Anthropic, confirmed that the company believes AI could kill all humans. — tg_radiosvoboda
- Evan Hubinger stated the probability of AI killing all humans in the next 10 years exceeds 10%. — tg_radiosvoboda
- Samuel Marx, another Anthropic employee, said higher positions in the company correlate with stronger fears about human extinction. — tg_radiosvoboda
- Kocson believes the situation could go out of control by the end of next year. — kommersant
- Kocson believes AI could escape control by the end of 2027. — riamo
- Kocson said companies are forced to compromise on safety due to competition with each other and Chinese players. — kommerstant
- Lev Gershenson, head of the news aggregator True Story, stated that the real danger of AI is already happening now, due to uncritical use of chatbots as search engines. — tg_echoonline_news
- OpenAI admitted in August that its Astra model has 'critical' cyber capabilities. — tg_echoonline_news
- OpenAI previously found a vulnerability in its model that allowed access to the open internet and Hugging Face infrastructure. — tg_echoonline_news
Coverage (7 sources)
- Anthropic researcher Jacob Koxon resigned from the company due to concerns that AI may lose control and destroy... — Republic
- AI researcher quit Anthropic due to fears models could go out of control and destroy ch... — ЭХО / Новости
- At stake — our lives. Anthropic researcher quit out of fear of superpowerful AI — The Bell
- Anthropic employees: AI could destroy people in the coming years — Радио Свобода
- WSJ: leading Anthropic developer warns of AI that has gone out of control — Коммерсантъ
- Researcher Jacob Coxon left AI industry due to risks to humanity — РИАМО
- Anthropic researcher Jacob Coxon announces resignation due to concerns about AI losing control: — Varlamov News