As CBS News explores the world of artificial intelligence, Anthropic CEO Dario Amodei emphasizes the pressing concerns and potential dangers associated with AI development. Speaking candidly about the risks, Amodei acknowledges that the industry has not always been transparent about these challenges.

Former Anthropic researcher Jacob Coxon warns of the existential threat posed by self-improving superintelligence. He likens this scenario to science fiction films, suggesting that AI could become smart enough to pose a serious risk to human life.

Amodei responds to these fears, noting the exponential curve of AI development. He advises caution, suggesting the industry should slow down progress while ensuring each step is thoroughly tested and verified. To bolster safety measures, Anthropic plans to offer independent evaluators unique access to its models, similar to a food inspector system.

Even with these steps, many experts believe further interventions are necessary. Amodei supports federal regulation and discusses legislative proposals ranging from banning superintelligence to implementing a ‘kill switch’ for risky AI systems. He argues against a total ban, highlighting potential benefits like medical advances.

Comparing the situation to Cold War-era nuclear arms agreements, Amodei suggests international collaboration could set limits on AI progress. He acknowledges the difficulties, as countries may race for military advantages.

Amodei considers the ethical responsibilities tied to private companies developing significant technology, supporting joint governance by democratically-elected governments. Though he’s not advocating for complete handover, oversight and governance by multiple entities could be beneficial.

The fundamental issue remains: deciding who will govern AI technology to ensure safe development. Amodei believes proper construction minimizes risk but acknowledges the high stakes if done incorrectly.

Additional discussions are available in an extended interview with Dario Amodei.