Why current AI models won’t make scientific breakthroughs: Thomas Wolf

A technology founding partner said that existing artificial intelligence models from laboratories such as Openai would lead to great scientific breakthroughs and poured cold water against the claims and claims of large figures in the field.

Thomas Wolf, the founding partner of the $ 4.5 billion AI starting face, is in danger, including Openai boss Sam Altman and anthropic CEO Dario Amodei with great names in AI.

When Wolf talks about scientific breakthroughs, it means new ideas such as Nobel Prize -level ones. Examples containing Nicolaus Copernicus, the theorizing the sun, were at the center of the universe and other planets move it.

Wolf has currently announced a few problems with Chatbots. The first is that these products, such as Chatgpt and others, often agree or align the same as the person who wants it. If you ask a chatbot a request, think again and will tell you how interesting or wonderful this problem is.

Secondly, the models that support these chat boots are designed for “the most likely to predict” or “word” in one sentence.

However, he noted the two main features of scientists. The first is that scientists who make great strides are often opposed and question what others say.

“The scientist does not try to predict the most likely word. In fact, this is not surprisingly, but in fact it is true that this is a lot of new things.” He said.

Huging Face Founding Partner has been thinking about it for the last few months. “AI-private biology and medicine, human biologists will allow us to reach the progress that will reach for 5-10 years in the next 50-100 years,” Stating that an article written by Antropic’s article was written by Amodei.

This allowed Wolf to think about how it would not be possible with the current model crop, in the AI’s state and its opinion.

Wolf said that these chat boots and vehicles will probably be used as a kind of “common pilot for a scientist” in which they are used to research to help people produce new ideas.

To some extent, that’s happening. Google Deepmind’s Alphafold product helped to analyze the protein structures that the company promised scientists to discover new drugs.

However, there are some new initiatives to take AI one step further to make scientific breakthroughs, including Lila Sciences and Futurehouse.