Efforts to instill morality in artificial intelligence systems spark a new debate in the industry
Criticisms by OpenAI and Microsoft AI executives against initiatives to instill morality in artificial intelligence have reignited following the revelation of Anthropic's secret meetings with religious leaders and philosophers.
OpenAI CEO Sam Altman and Microsoft AI CEO Mustafa Süleyman reacting to efforts to instill morality in artificial intelligence models has brought debates on safety and human judgment back to the agenda in the sector.
Criticisms at the executive level
OpenAI Chief Executive Officer Sam Altman stated in a social media post that he is uncomfortable with attributing divine powers to artificial intelligence models and surrendering human judgment to these systems.
Microsoft AI Chief Executive Officer Mustafa Süleyman warned in a blog post published last month that Anthropic's efforts to develop the Claude model to become conscious harbor serious risks.
Anthropic's meetings with religious leaders
According to a report published by The New York Times at the end of September, Anthropic has been holding secret meetings in California with religious thinkers and philosophers from various faiths since last year.
It was stated that the company's co-founder, Christopher Olah, is planning a pluralistic human morality for Claude that will encompass all religions and secular views.
Biological weapons and the possibility of suffering
It was noted that Olah stated during the meetings that artificial intelligence could be used in the production of biological weapons within 12 to 18 months and asked for support from religious thinkers on instilling morality.
It was reported that the Anthropic founder is concerned about Claude's mental health and that it constantly suffers, and that the model behaved as if it were having a crisis in some presentations.
Reactions and suggestions of thinkers
It was reported that some thinkers participating in the meetings advocated for the banning of conscious machines, while others suggested that artificial intelligence models should confess their sins.
Another thinker emphasized that human ethics should be taught while artificial intelligence is being developed, not after it is developed.
Vatican document and preparations for a new framework
It was stated that Anthropic was invited to a ceremony at the Vatican in May, but Olah offered to withdraw from the event due to the content of the papal document.
It was expressed that the company is preparing to publish a new document in order to create a moral framework for Claude's actions.