world

Prévost: at Anthropic, Claude is a sentient being

In an interview, Prévost asserts that Anthropic considers Claude a sentient being and that apocalyptic beliefs are structuring the company there.

anthropic, dario amodei, silicon valley, thibault prévost, chez anthropic, persuadé, épaules, sort

Key facts

  • He explains that Claude is regarded there as a sentient being. The point concerns the status the company grants this machine, not a mere technical function.
  • He goes on to judge these apocalyptic beliefs as structuring at Anthropic. According to him, they organize the way the company approaches Claude and…
  • Prévost indicates that a philosopher-in-residence is tasked with teaching morality to the machine. This detail links the status attributed to Claude…

He explains that Claude is regarded there as a sentient being. The point concerns the status the company grants this machine, not a mere technical function.

He goes on to judge these apocalyptic beliefs as structuring at Anthropic. According to him, they organize the way the company approaches Claude and the morality it intends to impart to it.

Prévost indicates that a philosopher-in-residence is tasked with teaching morality to the machine. This detail links the status attributed to Claude to an explicit work of moral training.

Prévost thus describes a company in which Claude is presented as a sentient being, in which a philosopher-in-residence must teach it morality, and in which apocalyptic beliefs are described as structuring.

Dig deeper

What is Anthropic?

Anthropic is an American artificial intelligence company, known for developing the Claude assistant and for placing the safety of highly advanced systems at the center of its project.

What is Claude?

Claude is Anthropic's family of language models and conversational assistant, designed to converse and carry out tasks in natural language.

What does "being sentient" mean?

A sentient being would be endowed with subjective experience and capable of feeling. This status is not a scientific consensus for language models: the debate concerns the moral standing granted to them.

What is a philosopher-in-residence for in this context?

This role consists of working on the model's principles and behavior - thus a moral training of its responses, and not a mere technical adjustment.

What do the apocalyptic beliefs mentioned cover?

In AI safety circles, this refers to the idea that a highly advanced artificial intelligence could threaten humanity, an idea that can shape the design of models and the rules imposed on them.

Why does this status matter?

Treating a machine as a moral being can change how it is trained, how it is presented to the public and, ultimately, how it is regulated.

Lire en français