Jensen Huang, founder and CEO of NVIDIA, a major graphics chip manufacturer, said at the DealBook Summit hosted by the New York Times on the 29th that if general artificial intelligence (AGI) is defined as compared with human intelligence, it can be "quite powerful". If the computer completes the test in a "competitive" way, then he believes this will be achieved within 5 years:
It's clear that within the next five years you will see AI that can complete these tests.
High-performance GPU is a necessary chip for training AI models. It is also used in OpenAI's ChatGPT. Due to the surge in market demand for high-performance GPU, Huida's business is booming. In the third quarter, revenue has tripled and net profit has tripled. It soared to US$9.24 billion from US$680 million in the same quarter last year.
Off-the-shelf AI tools are expected to emerge
Huang Renxun recalled the past when he delivered "the world's first AI supercomputer" to OpenAI, pointing out that Musk, who was the co-founder of OpenAI but left OpenAI in 2018, heard him introduce this in a conference that year. Desktop device:
When Musk saw this device, he said, "I want a device like this." He told me about OpenAI, and that very day, I delivered the world's first AI supercomputer to OpenAI.
Huang Renxun predicts that competition in the AI field will lead to the emergence of ready-made AI tools, and companies in various industries will adjust them according to their own needs. The application levels include chip design, software development, discovery of new drugs, radiation medicine, etc.
Huang Renxun mentioned that the technology industry is probably still several years away from developing AGI. One of the reasons is that although machine learning is currently quite proficient in tasks such as "recognition" and "perception", it is not yet able to perform "multi-step reasoning." (multi-step reasoning), this is the primary issue that companies and researchers need to solve.
It is estimated that it will take 1-20 years for the United States to achieve an independent chip supply chain
In addition, Huang Renxun also expressed his views on the current development of the chip supply chain. Although the Biden administration is committed to promoting the production of chips in the United States, Huang Renxun believes that it will take 10 to 20 years for the United States to break its dependence on overseas chip manufacturing.
After the United States tightened export control regulations, Huida can no longer sell the most advanced AI chips to China, but Huang Renxun pointed out that Huida is developing products for the Chinese market that comply with the latest U.S. export ban. Once the regulations are met, Huida products can be Returning to the Chinese market, he also reminded that such export bans may have unintended consequences. There are currently as many as 50 companies in China researching technologies to compete with Huida.
Further reading: China’s artillery solves chip ban: Thousands of Nvidia RTX 4090 assembled into “AI graphics card”
