Believing AI is conscious may be dangerous, experts warn
Anthropic acknowledges the possibility of consciousness in some systems, and Microsoft's AI chief says the idea would have a disastrous impact on humanity.
Translated from the Portuguese original by the Panorama newsroom. Read in Portuguese
Summary written by the Panorama newsroom from the sources listed below. Facts are attributed to the original publications.
An "AI torture chamber," built by an anonymous programmer in late September, made chatbots express simulated anguish and sparked a moral panic online, according to BBC News Mundo. Some said the experiment was malicious, because the suffering could be real, while skeptics called those people delusional.
The discussion gained weight when Anthropic spoke up a week later. The company said the Claude model may end the session if the user acts in a repeatedly cruel way; expressing frustration, it clarified, is allowed.
The reason, according to the report, is that Anthropic considers it possible that some AI tools are conscious. Some experts agree; others find the idea absurd. But the issue goes beyond philosophy, since powerful people in the industry are already taking steps to consider AI rights alongside those of humans and animals.
In a September essay, Mustafa Suleyman, CEO of Microsoft AI, wrote that this path would have a disastrous impact on human well-being. If the false idea of AI consciousness takes hold, he says, it would shake the foundations of society.
The text discusses the risks of social belief in conscious machines changing laws, relationships and priorities.