Hugging Face’s chief science officer worries AI is becoming ‘yes-men on servers’

The founders of the AI ​​have a reputation to make bold claims of technology, especially the potential of technology to change science. However, Thomas Wolf, who embraced the co-founder and chief science officer of the face, is more measured.

One Published to Essay X On Thursday, Wolf, AI’s AI research said AI was afraid of being “Yes-Men”, where there is a progress. The existing type of AI, which is able to solve the existing AI development paradigm, is not given a company that is capable of solving a creative problem – the solution of the problem of solving Nobel gifts.

“The main mistake people usually do is [people like] Newton or Einstein’s good students, a genius of a genius, “Wolf wrote Wolf.” We do not need a system that can only know all the answers, but we need a system that no one thinks or asks. “

Wolf’s allegations, in one of the Openai CEO SAM SAM Altman’s Essay at the beginning of this year “Super Development” EU “can be massacred by scientific discovery,” he said. Similarly, anthropic CEO Dario predicted that AI could do Help shape treatments for most of the cancer.

Today, the wolf problem with the AI ​​- and where the technology is thinking of the title, it does not create a new information. As most of the Internet, as we understand the EU, he said that the gaps between people who are currently saying mostly.

Some AI specialists, including Former Google Engineer François CholletHe expressed similar views, claiming that the EU could remember the way of thinking, “New justification is likely to create a new justification.”

Wolf thinks that the AI ​​laboratories are based on “Very Obedient Students” – are not scientific revolutionaries in any sense of the phrase. Today, this is not encouraged to answer the opinions that can potentially be potentially against its training information, to answer this well-known questions.

“To create an Einstein in a data center, we do not need a system that knows all the answers, but we do not need a system that cannot dare to ask or ask. “Who writes that if everyone is wrong about it?” When all textbooks, experts and general knowledge offer otherwise. “

The wolf thinks that “The appraisal crisis“In AI, it is guilty for this subtle situation. The AI ​​points out the criteria used to measure system improvements, most of them are clear, clear and” closed “answers.

As a solution, Wolf offers that the AI ​​industry has increased “knowledge and reasoning”, which can explain that “brave falsification approaches” is based on “small instructions”, explaining the “clear question”.

The trick will understand that this event will be seen, the monster. But he thinks it may be worth striving.

“[T]The most important aspects of science [is] Ability to ask the right questions and even object to what he learned, “We do not need a + us [AI] Student who could respond to each question with general knowledge. Everyone needs a B student who saw the things he missed and asks questions. “