Humanity is entering a phase of artificial intelligence development that will \u201ctest who we are as a species\u201d, the boss of the AI startup Anthropic has said, arguing that the world needs to \u201cwake up\u201d to the risks.<\/p>\n
Dario Amodei, a co-founder and the chief executive of the company behind the hit chatbot Claude<\/a>, voiced his fears in a 19,000-word essay<\/a> titled \u201cThe adolescence of technology\u201d.<\/p>\n Describing the arrival of highly powerful AI systems as potentially imminent, he wrote: \u201cI believe we are entering a rite of passage, both turbulent and inevitable, which will test who we are as a species.\u201d<\/p>\n Amodei added: \u201cHumanity is about to be handed almost unimaginable power, and it is deeply unclear whether our social, political, and technological systems possess the maturity to wield it.\u201d<\/p>\n The tech entrepreneur, whose company is reportedly worth $350bn<\/a> (\u00a3255bn), said his essay was an attempt to \u201cjolt people awake\u201d because the world needed to \u201cwake up\u201d to the need for action on AI safety.<\/p>\n Amodei published the text as the UK government announced Anthropic would help create chatbots that support jobseekers with career advice and finding employment, as part of developing an AI assistant for public services in general. Last week, the company published an 80-page \u201cconstitution\u201d for Claude in which it set out how it wanted to make its AI \u201cbroadly safe, broadly ethical\u201d.<\/p>\n Amodei co-founded Anthropic in 2021 along with other former staff members from OpenAI, which developed ChatGPT. A prominent voice for online safety known for warning consistently<\/a> of the dangers of unrestrained AI development, he wrote that the world was \u201cconsiderably closer to real danger\u201d in 2026 than it had been in 2023, when the debate over existential risk from AI raced up the political agenda<\/a>.<\/p>\n He alluded to the controversy over sexualised deepfakes created by Elon Musk\u2019s Grok AI that flooded the social media platform X<\/a> over Christmas and the new year, including warnings that the chatbot was creating child sexual abuse material<\/a>.<\/p>\n Amodei wrote: \u201cSome AI companies have shown a disturbing negligence towards the sexualisation of children in today\u2019s models, which makes me doubt that they\u2019ll show either the inclination or the ability to address autonomy risks in future models.\u201d<\/p>\n CEO and co-founder of Anthropic, Dario Amodei. Photograph: Denis Balibouse\/Reuters<\/p>\n The Anthropic CEO said powerful AI systems that could autonomously build their own systems could be as little as one to two years away.<\/p>\n He defined \u201cpowerful AI\u201d as a model that was smarter than a Nobel prizewinner across fields such as biology, mathematics, engineering and writing. It could give or take directions to or from humans, and although it \u201clived\u201d on a computer screen it could control robots and even design them for its own use.<\/p>\n While acknowledging that powerful AIs could be \u201cconsiderably further out\u201d than the two-year timeframe, Amodei said recent rapid progress made by the technology should be taken seriously.<\/p>\n \u201cIf the exponential continues \u2013 which is not certain, but now has a decade-long track record supporting it \u2013 then it cannot possibly be more than a few years before AI is better than humans at essentially everything,\u201d he wrote.<\/p>\n