AI Developer Claude Calls for Slowdown Amid High Risks
Translated from Russian, summarized and contextualized by DistantNews.
At a glance
- Anthropic, the developer of the AI model Claude, has called for a slowdown in AI development due to high risks.
- The company notes that AI's capabilities are doubling every four months, significantly faster than before.
- Anthropic proposes a pause in AI development to allow humanity to adapt and avoid losing control.
Anthropic, a leading AI developer known for its Claude model, is urging a pause in the rapid advancement of artificial intelligence, citing significant risks associated with the technology. The company's co-founder, Jack Clark, and head of research, Marina Favaro, highlighted the accelerating pace of AI development.
The facts indicate that the role of humans in the AI development process is becoming less significant with each step. As soon as the quality of code written by humans and artificial intelligence becomes equal, people will stop writing code and will only be engaged in its verification.
They observed that the scope of tasks AI can perform independently is doubling approximately every four months, a rate considerably faster than the previous seven-month interval. If this trend continues, AI models could soon handle tasks that currently take humans days or even weeks. Anthropic's experts believe human involvement in AI development is diminishing, with AI potentially surpassing human coding capabilities soon.
To address these concerns, Anthropic proposes a global "pause" on AI development, advocating for a period of adaptation for humanity. They warn that without such a pause, AI models might create their own successors, leading to a loss of human control. The company stresses the need for all major AI corporations, particularly in China and the United States, to support this initiative, as a unilateral pause by one company could be counterproductive if competitors accelerate their own development.
AI models will create their own analogues until we lose control over them.
Anthropic draws parallels to past international arms control treaties, like the Intermediate-Range Nuclear Forces Treaty, noting that establishing verification regimes for complex technologies took decades. They argue that humanity lacks such ample time to adapt to the rapid evolution of AI, emphasizing the urgency of their call for a slowdown.
Verification regimes for other complex technologies, such as the Intermediate-Range Nuclear Forces Treaty, took decades to establish, both in terms of infrastructure and trust. We do not have that much time.
Originally published by 24.kg in Russian. Translated, summarized, and contextualized by our editorial team with added local perspective. Read our editorial standards.