Jensen Huang, the CEO of Nvidia, believes that artificial ielligence does not give the best answers at the mome, and we are basically several years away from the release of artificial ielligence that can be trusted to give answers.
According to Business Insider, the CEO of Nvidia, who was prese at the Hong Kong University of Science and Technology, said that artificial ielligence answers are not the best answers at this mome. According to this technology manager, people should not have to doubt the answers they get from artificial ielligence and guess whether the answers correspond to reality or not.
Humanity is still several years away from providing reliable artificial ielligence
Jensen Huang coinued:
“We have to get to the poi where you have a lot of confidence in the answer you’re getting; So I think we are several years away from doing that, and at the same time we have to coinuously increase our calculations.”
Despite the advances in AI and the impressive developme of large-scale language models (LLM), it is still not possible to be 100% confide in AI answers. One of the biggest problems of this technology is the possibility of producing answers based on hallucinations. This problem has even been observed in OpenAI’s Whisper artificial ielligence, which is used for transcription.
Artificial ielligence companies may even face lawsuits for delirious chatbots; For example, last year a radio host sued OpenAI’s ChatGPT chatbot for quoting fake quotes from him.
The CEO of Nvidia believes that pre-training or training the model on a large and diverse dataset before developing it for a specific task is not enough. AI developers have yet to find a solution to the problem of their models delirious. Of course, Jensen Huang has a very positive view of the developme of artificial ielligence. He previously said that artificial ielligence can do some tasks a thousand times better than humans, but it cannot completely replace human labor.
