Fears of Science employees in the face of embrace, Amnesty International became “yes, a server men”

AI’s founders have a good reputation in making bold claims about technology capabilities to reshape the fields, especially science. But Thomas Wolf, co -founder of senior science officials, has a co -founder of science.

in An article published to X On Thursday, Wolf said he was afraid that Amnesty International will become “yes on servers” absent from the penetration of artificial intelligence research. It has been made clear that the current artificial intelligence development models will not result in Amnesty International capable of solving creative problems, a type of problem solving that wins the Nobel Awards.

Wolf wrote: “The main mistake that people usually make are thinking about (people who love) Newton or Einstein.” To create Einstein in a data center, we do not only need a system that knows all the answers, but it is something that can ask questions that no one else thought or dared to ask. “

The wolf’s assurances stand unlike those from the CEO of Openai Sam Altman, who is in Article earlier this year He said that “Superintyligent” can “accelerate a wide scientific discovery.” Likewise, the CEO of anthropologist Dario Amani predicted Help formulate treatments for most types of cancer.

The problem of the wolf with artificial intelligence today – and where it is believed that technology is heading – is that it does not generate any new knowledge by delivering facts that are not relevant previously. Wolf said that even with most of the Internet at its disposal, Amnesty International, as we now understand it, fill the gaps between what humans already know.

Some artificial intelligence experts, including Former engineer, François ShouletteSimilar views have expressed the pretext that although artificial intelligence may be able to preserve thinking patterns, it is unlikely to generate “new thinking” based on new situations.

Wolf believes that artificial intelligence laboratories adopt what is mainly “very obedient students” – not the scientific revolutionaries in any sense of this phrase. He said that artificial intelligence today does not motivate the question and suggest ideas that are likely to conflict with their training data, which limits the answer to the known questions.

“To create Einstein in a data center, we do not only need a system that knows all the answers, but he can ask questions that no one else thought or dared to ask.” “One writes” What if everyone is wrong in this? “When all textbooks, experts and common knowledge suggest that.”

The wolf believes that “”The evaluation crisis“In artificial intelligence is partially the cause of this annoying state. It refers to the criteria that are usually used to measure the improvements of the artificial intelligence system, most of which consist of questions that have clear, clear and” close “answers.

Kohl, Wolf suggested that the artificial intelligence industry “moves to a scale of knowledge and thinking” that is able to clarify whether AI can follow “bold adverse methods”, which provide general proposals based on “small hints”, and to ask “other situations” that lead to “new research paths”.

Wolf admits that the trick will recognize the form of this measure. But he believes he could deserve all this effort.

Wolf said: “(R) is the most important aspect of science (it is) the skill to ask the correct questions and challenge even what one has learned.” “We do not need a student A+ (AI) who can answer each question with general knowledge. We need a student B who sees and wondering what anyone has missed.”