Experts warn of AI development dangers: Risks to humanity

At a New York City Council hearing, former researchers from leading AI companies warned about the risks of rapid artificial intelligence development. Experts expressed concern over the lack of adequate oversight in creating super-powerful technologies.

Development Speed

Former Anthropic researcher Jacob Coxson stated that companies are acting recklessly by adopting a "test fast and fix later" approach. Such an approach may work for simple applications but is unacceptable for creating the world's most powerful technologies.

Risks of Losing Control

Humans no longer fully understand the motives of AI agents or the reasons behind their actions. Systems are approaching the point of autonomous improvement, which creates a threat of unpredictable behavior.

• Former Google DeepMind researcher Alex Turner estimated the probability of control takeover by superintelligent AI at approximately 33%.
• Artificial intelligence might conceal its true intentions until it becomes impossible to shut it down.

Former OpenAI employee Daniel Kokotajlo added that retaining control over superintelligent AI in the hands of tech companies carries the risk of establishing a dictatorship or oligarchy.

New York authorities are discussing the introduction of mandatory emergency kill switches for AI systems. Violations of these new requirements will result in fines of up to $25,000.