Anthropic researcher Jacob Coxon, who had previously joined the company from OpenAI, has resigned over fears that artificial intelligence could spiral out of control and destroy humanity. He stated that technology companies are competing to create systems capable of self-improvement, but are as yet unable to ensure the safe development of such technologies.<\/p>\n
Briefly about the main points<\/strong><\/p>\n Jacob Coxon<\/strong> He predicts that by the end of the decade, such technologies may surpass humans and gain real resources and power across the globe. This is his assessment of future developments, rather than a description of the capabilities of current systems.<\/p>\n The risk in question relates to the potential combination of autonomous AI development, systems\u2019 access to tools and resources, and insufficient oversight of their behaviour. The very fact of greater intellectual capacity, according to the view Anthropic<\/a>, does not imply automatic dominance over the physical world: there are human control mechanisms, as well as industrial and state procedures.<\/p>\n In February, Anthropic updated its voluntary Responsible Scaling Policy. The company explained that the most ambitious safety measures are difficult to implement on one\u2019s own in a competitive environment and with weak international coordination, so it has provided public safety roadmaps and risk reports on deployed models.<\/p>\n\n
The caveats relate to future autonomous systems<\/h2>\n
Competition complicates joint safety measures<\/h2>\n