Anthropic urges debate on slowing AI development
Anthropic, the developer of the artificial intelligence assistant Claude, has stated that the option of slowing down or temporarily pausing AI development should be considered.
The AI company warned that artificial intelligence systems are increasingly capable of carrying out software development and research processes with less human involvement. If this trend continues, such systems could eventually evolve into technologies capable of improving themselves, CE Report quotes Anadolu Agency.
Anthropic published a blog post titled "When AI Builds AI," written by co-founder Jack Clark and Anthropic Institute head Marina Favaro.
The authors noted that while AI development was historically carried out largely by humans, an increasing share of the process at Anthropic is now being delegated to AI systems, accelerating the pace of development.
According to the blog post, if this trend continues, AI systems may eventually reach a stage where they can design and develop their own successors — a concept known as "recursive self-improvement."
“We are not there yet, and recursive self-improvement is not inevitable. However, it could happen sooner than most institutions are prepared for,” the authors wrote.
The post stated that current technical trends suggest AI systems are likely to become significantly more capable in the coming years. While self-improving AI could generate enormous benefits in science, healthcare, and many other fields, it could also increase the risk of humans losing control over AI systems.
“If systems are capable of fully building their successors, our methods for securing, monitoring, and shaping their behavior become far more important,” the authors said.
The blog further suggested that slowing down or temporarily pausing advanced AI development may be worth considering in order to give social institutions time to adapt to technological progress.
According to Anthropic, such measures would be most meaningful if implemented simultaneously by multiple leading AI laboratories across several countries under similar conditions.
Photo: X
