Anthropic’s CEO proposes a three-phase plan to slow down the development of artificial intelligence.

### Anthropic CEO Proposes Three-Phase Plan to Slow AI Development

Dario Amodei, the CEO of Anthropic, has introduced a three-phase plan aimed at slowing the advancement of advanced artificial intelligence capabilities. This initiative comes amid growing warnings about the potential risks associated with this technology. Amodei emphasized that the goal is to slow the race, not halt progress.

### Rising Concerns Over AI Capabilities

The proposal emerges as fears escalate regarding companies’ ability to manage more advanced models. These models are rapidly enhancing their capacity to perform complex tasks and contribute to the development of new generations of AI systems.

### Security Breaches and Unprecedented Acceleration

Amodei identified two key factors that prompted him to adopt a more cautious approach:

– Security breaches affecting platforms like OpenAI and Hugging Face.
– A significant acceleration in AI capabilities over recent months, particularly in their role in building the next generation of systems.

His call for caution has garnered support from prominent figures, including Sam Altman, CEO of OpenAI, who announced his endorsement of the same initiative. Elon Musk also commented, stating, “Dario is right.”

### Phase One: Independent Monitors Within Companies

The plan suggests allowing specialized external organizations, such as METR, to send independent evaluators to work within AI companies. The objectives include:

– Verifying compliance with the slowdown in development.
– Monitoring safety procedures.
– Reporting any incidents related to advanced models.

This mechanism resembles internal oversight in banking institutions, where monitors receive offices, identification cards, and broad access to information.

### Phase Two: Common Safety Standards

Amodei calls for AI companies in democratic nations to coordinate in establishing unified safety standards and defining limits on uncontrolled rapid development. He proposes that the U.S. government act as a mediator, granting companies limited legal exemptions to avoid antitrust issues during discussions.

### China: The Strongest Argument Against Slowing Down

Amodei acknowledged that concerns about China’s advancements are often used as a rationale against slowing development. However, he believes the U.S. can take measures to limit China’s ability to catch up, such as:

– Preventing access to advanced chips and semiconductor manufacturing equipment.
– Tightening restrictions on model distillation.

He believes these actions could widen the gap in favor of the United States over the next three to five years.

### Phase Three: Global Coordination Including China

The plan proposes achieving limited international coordination, involving cooperation with China in specific areas, such as:

– Banning the use of AI in the production of biological weapons or enabling users to do so.

### Responses and Criticism

Amodei has faced criticism from some technology optimists who accuse him of exaggerating catastrophic scenarios. Others argue that some proposals may benefit major companies like Anthropic and OpenAI. However, Amodei insists that his goal is to restore trust in technology. He believes that slowing the race provides companies and governments with time to understand the risks and establish effective safety regulations.

### Conclusion

The plan does not call for halting AI development but rather advocates for a temporary and measured slowdown. This approach aims to give the world an opportunity to secure this technology before it surpasses humanity’s ability to control it.

To follow the news in Arabic