"We believe that artificial intelligence is not a fleeting trend, nor is it an inflated bubble," said Yang Yuanqing, Chairman of Lenovo Group, at today's Lenovo Tech World conference.

He cited this year's Nobel Prizes as an example, noting that the selection process is very cautious and often requires years of verification. The fact that this year both the Physics and Chemistry prizes were awarded to scientists related to AI further illustrates that AI is not a bubble or a fiction.

However, he also acknowledged that the development of AI will definitely go through a tortuous process. For instance, the demand for a certain AI product may suddenly surge at a certain stage and then plummet. But the overall direction of AI is determined; it has already improved the quality of life for individuals and provided higher productivity for businesses.

Advertisement

Nevertheless, in his view, the essence of AI is not just about large language models, nor is it solely about so-called GPUs or AI servers. In the digital age, every person and every company has accumulated a large amount of data. "The essence of the intelligent era is how to make these data valuable and intelligent. That's the essence of intelligence."

He gave an example that whether it's an individual or a company, they all have their own exclusive and proprietary data. You may not want to share it with others, or there may be no need to share it. But based on it, it can help users generate intelligence, making work and life more convenient and efficient. For companies, it's about optimizing processes and improving efficiency, such as increasing productivity and the accuracy of decision-making. "So we believe AI will not be a bubble."

Does he believe in AGI (Artificial General Intelligence)?

Lenovo's new Chief Technology Officer, Tolga Kurtoglu, and Yang Yuanqing agree that AGI is still a distant goal.

Although AI is making progress every day and has achieved significant progress in certain specialized fields, "just like 10 years ago when AlphaGo came out, it was hard to imagine that machines would defeat human Go players. When Deep Blue came out, you couldn't believe that humans would lose to machines. So now, in certain specific areas, we will continue to see breakthroughs, but to completely replace humans, to think like humans, to make decisions and act like humans, there is still a long way to go," said Yang Yuanqing.

He also mentioned that the industry's large models have made breakthroughs in text-to-image and even video, but these are all based on two-dimensional levels. As for how to enter the three-dimensional world, some people have already started to tackle it, including the spatial intelligence that Li Feifei has been working on in her startup. But this is more challenging. Because language intelligence has a large amount of data that can be trained, in the three-dimensional field, except for the field of intelligent driving, which has already received sufficient training data, there are not many other scenarios. "Including robots, if you ask it to pick up water, it doesn't know the weight, and the water in the cup spills out, but humans won't."

And how will AI change the future of PCs? If we look at a cycle of 10 or 20 years, will the form of computers in the future be very different from today?Yang Yuanqing painted a picture for reporters of the possible changes in the form of computers over the next 10-20 years. He proposed the concept of "artificial intelligence twins," where everyone might have a virtual "twin" that generates intelligence based on their own behavioral data. This virtual twin could act on your behalf like a human, such as making credit card payments or booking flights, while the user simply inputs their requests. This would represent a significant change in the way computers are used in the future.

On the hardware level, he believes that any technology that aligns with natural human usage habits, such as voice, text input, and keyboards, will continue to exist. People will choose their preferred input methods based on personal preference, but some less commonly used technologies may gradually disappear.

Tolga added that all devices are essentially interfaces for people to interact with information. Current computer device designs are primarily based on command mechanisms and graphical user interfaces, initially intended for processing text information. As the trend towards multimodal information develops, users will need to not only handle text but also express and create information more naturally, such as through voice. The industry has reason to anticipate the emergence of new interaction methods and device forms. This means that future hardware design will no longer be limited to existing forms but will evolve with changes in information processing and computational demands, meeting users' diverse needs in a more natural and intuitive way.