Ever since ChatGPT took the world by storm, AI large models have become the darling of venture capital. However, after nearly two years of development, AI large models are still in a technological race, with little to show for in terms of revenue.
On October 8th, at the Peking University & AI Unicorn Summit, Yan Yan, Chief Partner of the SAIF Investment Fund and a member of the Investment Committee of the Peking University Foundation, stated, "AI should be the largest technological and social revolution in human history because it will change the way we behave. But it will inevitably come with a huge market bubble; as far as I can see now, it's all bubbles."
Although AI large models are still in their early stages and have not yet developed a sustainable and profitable business model, Yan Yan is very optimistic about the future prospects of AI. He believes that the opportunities brought by AI will be greater than those of the internet and is very certain that any symbolic processing or repetitive work will be replaced in the future.
In his view, in the B2B sector, the application of AI large models should be the most extensive, including software engineers, marketing, finance, pharmaceuticals, and other scenarios where AI will be present. As for the B2C side, application scenarios include personal assistants and education and training. Taking education and training as an example, many professors may become unemployed in the future, and the best courses internationally will be available online.
Advertisement
Regarding the employment or entrepreneurial choices of young people, Yan Yan believes that all jobs related to emotional creation are safe. If one wants to start a business, they can also engage in emotional-related work or creative endeavors, which would be safer.
The judgment behind the above mainly comes from Yan Yan's understanding of AI large models themselves.
Yan Yan stated that AI has been around for a long time. In 1956, McCarthy, Minsky, Rochester, and Shannon jointly initiated the "Dartmouth Summer Conference," where "Artificial Intelligence" was first proposed as a new discipline. In 2017, Google published a particularly important paper on the Transformer large language model, which did two things: one was the vectorization of concepts, and the other was relationships, combining multiple relationships for parallel processing.
This directly led to the emergence of high-order correlation problems. Yan Yan further explained, "Now, ChatGPT4 has about 1.5 million tokens, and it can form about 1.8 trillion parameters. Human cognition has been univariate so far, with only a few smart people reaching the second order. Transformer can feed a book into it and know the relationship between every word and another word, which is basically beyond human intelligence."
To a large extent, it is precisely because of the intelligent attributes of AI large models that they can be effective in all aspects of social life, and even AI large models are referred to as the Fourth Industrial Revolution. However, it is a product of rationality and logic, and ChatGPT has not had subjective consciousness to this day.
Yan Yan stated that ChatGPT has no subject or intention, while humans have intentions and emotions, so today's computer artificial intelligence is best suited to be our assistant. "In the future, when it comes to B2C, the largest application of computer applications will be the digital super twin, that is, the digital self."In other words, the biggest difference between AI large models and humans is the lack of self.
Leave a Comment