By Ryan Welsh, Kyndi™ Founder & CEO
Recently there have been increasing assertions that China will become the world’s artificial intelligence superpower because it has the most data. Most recently, Kai-Fu Lee was on 60 Minutes asserting this by saying “China’s advantage is in the amount of data it collects.” He’s wrong.
First and foremost, Dr. Lee is making a confident statement based on an assumption that has a very low probability of being correct. The “AI” Dr. Lee is referring to is deep learning. He assumes that deep learning has no limits and all by itself will get us to artificial general intelligence if we train it on more data. At present, though, there is more evidence to the contrary than there is evidence supporting this assumption. While deep learning performs very well on specific tasks, mainly perceptual classification like image recognition (provided that the image does not have too much clutter, however), it is not yet up to the task of handling the complexity of natural language, and numerous other functions that the average person performs daily. But sweeping these limitations under the rug has become the norm in the deep learning community, to the detriment of the AI industry as a whole.
Given deep learning’s current notable limitations, to be the AI superpower, China must invent new AI methods. However, scientific progress and authoritarianism are at odds. At its core, science is about discovering basic truths about the physical and natural world while authoritarianism is about manipulating or creating an alternative reality to concentrate power and limit political freedom. Think this point doesn’t matter because the Communist Party of China has loosened its grip? Watch Dr. Lee become “uncharacteristically shy” (08:15) when asked to comment on statements from President Xi Jinping. The Chinese political system does not encourage open exploration and examination of scientific truths. Until that changes, the best China can do is keep up with the world’s AI powers, not overtake them.
Third, scientific research involves collaboration among scientists, and increasingly across international borders. However, as China ascended into the upper echelons of global economic power over the past 30 years, it has done so by sacrificing relationships that it now needs to become the AI superpower. More and more people and organizations don’t want to collaborate with Chinese organizations. Ask any Silicon Valley startup if they would take venture investment or partner with China’s largest companies, and by and large, the answer is “no.” Why? Fear that those companies will steal their intellectual property. Ask any researcher if a researcher in China is copying their work, and the likely answer is “yes.” In a July 2013 article in The New York Times, a representative from the University of Wisconsin stated that the university received “90,000 to 100,000 attempts per day, from China alone, to penetrate [their] system.” Would you collaborate with someone who consistently tries to take from you?
While China has done a great job using deep learning methods in novel ways, Dr. Lee vastly overstates the inevitability of China becoming the world’s AI superpower. I hope all this changes, because AI has the power to change the world for the better more than anything in the history of mankind. And that’s the one thing Dr. Lee and I agree on.