“It won’t be fast. Nvidia CEO predicted when AI will get rid of hallucinations November 26, 08:33 Share: Hallucinations remain a serious AI problem (Photo: Screenshot from Bipartisan Policy Center / YouTube video) Author: Anastasia Pechenyuk Nvidia CEO Jensen Huang is convinced that solving problems with whimsical and incorrect AI responses, often called hallucinations, will be possible after significant increase in computing power.”, — write on: ua.news
Hallucinations remain a serious problem in AI (Photo: Screenshot from video Bipartisan Policy Center / YouTube)
Nvidia CEO Jensen Huang is convinced that solving problems with bizarre and incorrect answers of artificial intelligence, often called hallucinations, will be possible after a significant increase in computing power.
The artificial intelligence industry will not be able to solve the problem of artificial intelligence hallucinations for several more years, believes the CEO of Nvidia, the world’s largest supplier of AI chips, Jensen Huang.
“Today, the answers we have are the best possible. But we have to get to the point where the answer we get is not the best possible answer, and we have to decide to some extent whether it is a hallucination or not a hallucination, whether it makes sense, whether it is reasonable or not. We must come to the point where the answer we receive is largely trusted. I think it will be a few more years before we can do that. In the meantime, we must continue to scale up our computing,” he said during an event at the Hong Kong University of Science and Technology (HKUST).
Currently, hallucinations, generated responses that contain false or misleading information presented as fact, remain one of the main problems with AI chatbots. This year, researchers from Google DeepMind and Stanford University proposed a method to combat them. They introduced a new automated scoring system, SAFE, which analyzes answers provided by artificial intelligence and looks for false ones.