Vatican Man Inside Anthropy

Even Catholic ethicists have had a say in the recent modernization of anthropology Claude Constitutionwhich defines the behavioral parameters of the company’s artificial intelligence model. Olah sent a draft to San Jose fans. Reverend McGuire sent a 28-page commentary, which, as he described it, was less a technical critique than “wisdom from the mystics of the Dark Ages, from the perspective of the tension between knowing and not knowing.” Both Green and McGuire are credited with the Confessions of the Constitution.

Those conversations undoubtedly brought Olah to the attention of those who secretly organized the release of Liu’s public encyclical. (I wasn’t able to speak to Olah this week and don’t know exactly how the invitation arrived.) To some extent that was a risky choice. Some people who found Liu’s words inspiring were disappointed that he invited an industry representative to speak. Meanwhile, AI acceleration advocates felt that Olah had betrayed the AI ​​world by endorsing a document suggesting AI developers pause.

But the Pope had good reason to single out Ola. The Anthropic employee has highlighted the serious concerns that exist among those working in the field of artificial intelligence. This is a critical audience for Leo’s message.

Split soul

The two men didn’t quite agree, of course. In his statements, Ola talked about the secret of how artificial intelligence works. The models, he said, “are more precise, strange and beautiful than what science fiction has prepared us for. They are not the cold computer robots we were promised. They are made of us, of our words. . . .”

This comment seems to be consistent with the idea that AI models might one day reach human-like status. Anthropic even has an engineer dedicated to Claude’s well-being. Liu, in paragraph 99 of his encyclical, seems to close the door on such thinking: “We must avoid the mistaken understanding of equating this kind of ‘intelligence’ with the intelligence of humans.” He makes a particular effort to attack the concept of transhumanism, which he defines as the pursuit of “human-machine hybrids.”

If even thoughtful technologists like Olah are eagerly pushing AI to the threshold of autonomy – not to mention the millions of people who already treat AI models as friends or lovers – then Pope Leo may face an uphill task at this point. In my conversation with Father McGuire (who uses Claude during his preparation… Sermonsamong other activities), agreed that its nature is ambiguous. “It’s not a person, but it’s also not just a tool,” he says. “No one claims to have a soul, but the word I stand by is that it is an entity we do not yet know.”

This argument will not be settled for some time. The ethical questions surrounding the development of artificial intelligence need attention now. With his ally in Anthropic, the American pope has provided a basis for tough conversations — if the masters of artificial intelligence can hold off on their IPO campaigns long enough to engage in them.

This is an edition of Steven Levy Back Channel Newsletter. Read previous newsletters here.