Anthropic CEO Dario Amodei has published a proposal for a controlled slowdown in current AI development in a blog post. In his statement for the Science Media Center, Prof. Dr. Matthias Kettemann explained that the proposal should be taken more seriously than earlier calls for a “pause” because it is not a vague call for a halt.
However, Kettemann also says the proposal should be read critically because: “We are dealing with a company whose business model would not be significantly impacted by a slowdown, yet it is simultaneously calling for export controls against Chinese competitors. The prediction that a swarm of agents could ‘take over the entire Internet’ in six to twelve months is unsubstantiated and cannot be substantiated.”
Statements by Matthias Kettemann and other researchers are available in German on the Science Media Center website.
Background
Following several security incidents involving autonomous AI systems that revealed inadequate safety measures, Dario Amodei, the CEO of Anthropic – the company behind the chatbot Claude – called for better industry-wide oversight and a slower pace of development for powerful AI models. These measures are intended to give providers time to implement them. Other major U.S. AI providers, such as OpenAI and Google, agreed with Amodei on many points.
The Science Media Center asked scientists to evaluate the current situation and potential future risks.