The Future Pioneered by Generative AI: Tokyo University × Generative AI Symposium "The Future Pioneered by Generative AI and Japan's Outlook"
A panel discussion featuring prominent figures in the field of AI took place at the University of Tokyo.
Chairman Masayoshi Son's perspective on AI was thought-provoking, so it is introduced below.
The Future of AI According to Chairman Masayoshi Son
Take a moment to imagine a broader picture.
Consider the comparison between an average human and the smartest fish. No matter how well-trained the fish might be, I believe it cannot match the brain activity of an average human. Some might argue that there are exceptionally intelligent fish, but let me pose another question. When comparing the intelligence of highly trained butterflies with that of an average human, who do you think is smarter? I doubt anyone would claim that butterflies are more intelligent or that they can use tools, speak, or operate computers.
In terms of the number of neurons, the human brain has approximately ten thousand times more neurons than a fish. The gap between butterflies and humans is around four hundred thousand times. It's a difference of about a million times. Generally, the number of neurons correlates with the wisdom and knowledge possessed by a living being. A complex and numerous neural structure allows for learning and thinking.
As of now, GPT4 is capable of excelling in various university entrance exams in the United States, such as medicine, law, mathematics, physics, and science, ranking in the top 10%. In other words, GPT4 can pass the entrance exams of most average humans. Personally, I have no doubt that I cannot simultaneously pass exams in medicine, law, and other subjects. I don't possess the knowledge to pass all those subjects, but I think an average human might.
If GPT4 has the knowledge equivalent to a score of 1, I believe that in the next ten years, it will increase its capacity by more than a million times. So, from the current level where GPT4 is already on par with an average human in terms of entrance exams, it will advance to a level equal to or even surpassing an average human, and this will happen in less than ten years.
I can assert this because I have been closely observing various developments. If AI surpasses humans by a factor of one million within the next ten years, and then further multiplies that gap by around a thousand times in the following decade, it will reach a difference of one trillion times. When there's a gap of a million times and a thousand times more than that, it becomes utterly futile to debate whether AI is smarter than humans in any field.
Until now, humans have been the most intelligent creatures on Earth, but that era will come to an end. At that time, we will need to reevaluate what happiness and human work mean, as well as reconsider harmony and peace – fundamental themes that will require intense discussion.
Currently, GPT's disclaimer is that it has only been trained until September 2021, which means it cannot provide comments on anything beyond that point. However, there will come a time when GPT is trained in real-time, with significant news, stock prices, weather, and the like. When the number of chips in GPT's platform exceeds one million times its current capacity, we will be stepping into the world of AGI (Artificial General Intelligence).
AGI could be likened to a crystal ball. If you ask a crystal ball about tomorrow's events or how to resolve current problems, it will provide accurate answers to most inquiries. If such an AGI world arrives within the next ten years, it will be applicable to almost everything. Instead of debating what it can be applied to, we will be in a beneficial era where it can be utilized for nearly everything. That's how I see it.
Ryusei Kakujo
Focusing on data science for mobility
Bench Press 100kg!