AI researchers warn of risk to humanity

Several AI researchers from leading companies have this week expressed serious concerns about the safety of advanced artificial intelligence. A researcher at Anthropic says there is a more than 10% chance that AI could lead to the death of all humans within the next decade.

In brief:

• Researchers at Anthropic, OpenAI and Google DeepMind warn of existential AI risks
• One researcher says AI companies are acting recklessly and working directly towards superintelligence
• According to several researchers, there are currently no plans to solve alignment problems

Researcher leaves Anthropic

A researcher who spent three years working on pre-training research at both OpenAI and Anthropic left Anthropic this week. He says that neither company is acting responsibly and that they are moving at breakneck speed towards self-improving superintelligence.

His criticism was confirmed by colleagues at the same companies. A researcher working on AI alignment at Anthropic confirms that the team genuinely believes AI could kill everyone on Earth, and puts the chance of this happening at more than 10% within ten years.

Several researchers voice concerns

Another researcher, who left Google DeepMind in June, fully supported these warnings. According to him, many researchers believe they are working on something that could kill everyone on the planet. This was literally part of his daily work.

The joint warnings suggest that there are serious doubts within the AI industry about the ability to keep advanced systems safe. Researchers point out that although Anthropic is doing its utmost, there is currently no plan to solve alignment problems for superintelligence, and that clearly no progress is being made.