Kai-Fu Lee has been studying the impact of artificial intelligence for decades. Kai-Fu Lee was a former Google executive who later became a venture capitalist. In his 2018 book, Artificial Intelligence Superpowers, he studied U.S.-China competition. His new book, AI 2041: Ten Imaginations for the Future, co-authored with science fiction writer Chen Qifan, proposes how artificial intelligence will bring earth-shaking changes to people's daily lives in the next 20 years.
Kai-Fu Lee currently manages $2.7 billion in assets at the Beijing-based Innovation Factory. Innovation Factory has invested in seven AI startups with a market capitalization of more than $1 billion: AInnovation, 4Paradigm, Megvii, Momenta, WeRide, Horizon Robotics and Bitman.
A reporter conducted an exclusive interview with Kai-Fu Lee, and some of the interviews were as follows:
Reporter: Before we discuss this new book, let's review "The Superpowers of Artificial Intelligence." How has the U.S.-China competition unfolded since its publication three years ago?
Kai-Fu Lee: Both countries continue to be what I predict as "AI superpowers," far ahead of the world. The U.S. is stronger in basic research, enterprise software, and enterprise artificial intelligence. China has accelerated the development of robotics and manufacturing. Pretty good match, right? The U.S. is stronger in enterprise software and China is stronger in manufacturing. Naturally, each will apply AI to the areas it excels at. Both countries have done a lot of work in medical artificial intelligence. It's a sudden occurrence that goes beyond my prediction of "AI superpowers" because COVID-19 has really accelerated a lot. This has led to AI-based drug discovery, protein folding, intelligence-based radiology, AI for social distancing, and the use of robots in the lab. The lab proves to be a great place for robots.
One of the portfolio companies we invest in is MegaRobo. The company has developed a robot that does COVID-19 tests 100 times faster than it did without a machine because it's all-weather and everything is in parallel. The same product can be extended to laboratory research and essentially replace the work done by many lab technicians. We've seen a lot of big breakthroughs in healthcare.
In terms of basic research, the total number of papers in China has caught up with the United States, and the number of papers also ranks in the top 50%. It almost caught up with the top 10 percent of papers; it still lagged behind in the top 1 percent. One theory is that young researchers in China are very similar, if not more, to AI researchers in the United States. China still lacks more senior and talented researchers, but young researchers are expected to fill the gap in the next 10 or 20 years.
So overall, I think both countries are doing a good job. If you look a year or a half further, counting the number of AI IPOs, China is actually leading. By the end of next year, 10 to 12 AI companies with multibillion-dollar valuations are expected to go public in China. I don't think the numbers in the U.S. would be that big.
Reporter: The format of "AI 2041" is different from "Artificial Intelligence SuperPower", it uses fictional stories. How did you react?
Kai-Fu Lee: Basically very positive. I want to use imaginative stories to introduce AI to people who might find AI intimidating. The next 20 years are a good time – any time too short will not seem revolutionary; any time too long is not a reliable prediction. The stories take place in 10 different countries and affect 10 different industries.
Reporter: If you had to choose the most important one from 10 options, which would you choose and why?
Kai-Fu Lee: "Contactless love" will be the most important because it illustrates the power of AI in healthcare. The COVID-19 pandemic has made us all realize how fragile humanity is. AI is ready to help cope with future epidemics, future health, and future healthcare. In areas such as wearable computers, radiology, gene sequencing, and more detailed blood testing, human healthcare is digitizing. Whether it's diagnosis, treatment, preventive maintenance, or health checks, data can be fed into an AI engine. An AI medical assistant is showing us what our weaknesses are, how to reduce them, what kind of medicines to take, and how to exercise in a way that suits us personally. When there is a lot of data, AI is well suited for high-precision recommendations and decisions. Many other AI technologies have enormous implications, but they can be used for the good as well as for the bad. One area of technology that I think is really "AI beneficial" is healthcare.
Reporter: Let's relate this book to your investment in the Innovation Factory. First, how important is AI in your portfolio?
Kai-Fu Lee: This is important and probably related to 70% of our investment. They're not all pure AI companies. Advances in artificial intelligence have made it possible for some proven technologies to still be injected into products and industries. The real breakthrough is 10% or 15% of our investment. The others are what I think are AI applications. Overall, it's about 70 percent.
Reporter: Can you tell us about some of the investments you've made? How do they fit into your vision in the book?
Kai-Fu Lee: The chapter on "Deep Fraud" is about computer perception and vision. Much of this goes back to deep learning and convolutional neural networks, both of which became popular around 5 to 7 years ago. What we see in the innovation workshop is that computer vision will be disruptive, and machines will be able to see things like people – not exactly the same, but like people. For example, it does a better job than people when it comes to recognizing objects and faces. We found that this could be used for visual inspection of quality in autonomous retail stores, radiology and manufacturing. It can also be used for navigation robots and autonomous vehicles. Deep learning can be applied to computer vision, and it will have many industrial applications. In the field of computer vision applications, we have a total of 14 companies. Several of these companies will go public by the end of next year.
Another breakthrough we are seeing now is a large, self-supervising natural language model that can be fine-tuned. I think natural language will be the next big AI application. Vision comes first, followed by natural language. You can see in the book "Twin Sparrow" how AI trains children's peers, can converse fluently in human language, can have very coherent conversations with children, and help them move forward. So far, we've made four investments in the natural language space, and probably a dozen more to create a portfolio similar to computer vision.
Reporter: What is the future of the entertainment industry?
Kai-Fu Lee: We didn't invest in that area because I think it's a bit far away. AR-VR is still too clumsy. The head-mounted display is too big. People get dizzy. Content production costs are too high. I think over time, these issues will be solved, the gaming experience will be truly vivid and realistic, it won't be dizzying, and the cost of producing content will be relatively low. This will be a time span of 5 to 10 years.
Reporter: How will autonomous driving develop?
Kai-Fu Lee: We see transportation as a huge revolutionary area for ai-powered self-driving cars. However, if you think it completely replaces the human driver and removes the steering wheel, accelerator and brakes from the car, then we need at least another 10 years, if not 15 years. It's hard to make a successful investment in this area because you have to keep burning money and collecting data for a long time without any gain.
A more appropriate approach is to find fruit that is easy to pick. A more constrained environment, a simpler problem, where you can collect data and gradually make your AI work better. We have invested in four companies that do this. One is airport transportation, and the other is port transportation – handling crates. The third is to do after-sales products, help drivers and collect data. The last company is making micro-robot buses — which they will deploy in three cities this year. These robotic minibuses are easier to deploy than robotic cars and robo-taxis because robot buses have fixed routes. Once you've collected this data for the robot bus, then you can get the robot taxi. For each, we don't need to wait 10-15 years to get paid.
At the end of the book, you think that the development of artificial intelligence will challenge traditional economic theory. Can you talk about this?
Kai-Fu Lee: If we look at the period from 20 to 40 years after 2041, I think the cost of goods will fall sharply. Labor costs will be reduced due to automation. With advances in life sciences and materials science, the cost of materials will also be greatly reduced. Distributed solar and battery storage will also reduce energy costs. Over the past 10 years, we have seen tremendous progress.
It's not just our mobile phones or everyday goods like cars that cost them should come down. What is not currently manufactured will become manufacturable. For example, food, vegetables and meat will start being produced in factories, not necessarily on farms. We will boil down many of today's problems to manufacturing. At this stage, the cost of providing housing, food, basic transportation and clothing for everyone should be feasible. That is, we should be able to eradicate poverty and hunger. At the same time, our existing economic system will no longer be effective. This includes a market economy or capitalism, the stock market, money and jobs. In the book, I call it the "Age of Plenty."
In the future, daily work will be replaced by artificial intelligence, so the meaning of work may change. The meaning of money may change because the state may want to subsidize everyone with a universal basic income. We may also want to motivate people, not just money; Money may not be the only measure of compensation.
I think the positive aspect is that we are able to produce a lot of things at a lower price. Artificial intelligence will do all the day-to-day work for us, but what if jobs are replaced? What about companies that need profits to justify their stock prices? These are all issues that must be addressed. For every benefit, there are also a range of issues that need to be addressed.