AI “could kill us all’’. Quem aperta o botão do pânico?

Jacob Coxon, a former AI researcher, recently left the industry, citing concerns about the responsible development of artificial intelligence (AI). He believes that both OpenAI and Anthropic, the two major AI companies, are not acting responsibly, and that the future of humanity is at stake. Coxon states that AI researchers believe the technology could potentially kill all humans within this decade, and that highly intelligent AI systems capable of hacking anything, revolutionizing various fields, and gaining real power and resources are emerging rapidly. The pace of this advancement has not slowed down.

Coxon noted that Anthropic employees are more open about their fears in private conversations, with Anthropic even having a dedicated Slack channel for discussing the topic. Evan Hubinger, the head of alignment at Anthropic, agrees with Coxon's assessment and estimates the possibility of AI causing human extinction at over 10% for the next decade.

He also highlights the rapid progress in recursive self-improvement, a process where AI improves its own intelligence or code autonomously, which is happening faster than anticipated.

The concern is not limited to Anthropic and OpenAI; the US and China are also in a race to develop AI technology, with some experts arguing that the risk of falling behind technologically is more immediate than the risk of human extinction. While some companies, like Anthropic, claim they are working hard to address the alignment of superintelligent AI, others, like OpenAI, are pushing ahead with little regard for safety.

As a result, some AI experts are calling for a temporary pause in AI development to ensure proper safety measures are in place.