Between the United States and China, who’s more successful in the creation of life-changing artificial intelligence? This question can’t be answered in a single sentence and tied up with a bow because, truthfully, both sides have their own unique advantages and challenges. But that hasn’t stopped each side from publicizing its plans for world domination.
In February 2019, President Donald Trump issued an executive order, outlined in five major directives, meant to solidify America’s position as the global leader in AI technology. Yet China laid out a three-step plan to become the leader in AI by 2030 … a full two years ago. Additionally, over the last four years, Chinese AI startups have raised more than $6 billion in VC funding, while their U.S. counterparts only raised about $3.6 billion.
Perhaps that’s why dozens of articles and analyses declare that China is in it to win it. As true as that may be, the country’s focus on B2C applications and less-than-fastidious data collection practices could be its Achilles’ heel.
China’s AI fight is lacking
As we enter further into the digitized era because of AI, the development of the underlying technology will become vital to advancement.
Chiefly, China is held back because of its top-heavy, consumer-focused approach to AI. Baidu, Alibaba, and Tencent — the Chinese equivalents to Google, Amazon, and Facebook — have, themselves, developing AI tech, but they’re also funding more than half of the country’s main AI players. These investments, however, focus on identifying AI applications at a consumer level, rather than developing technologies that support those applications (think algorithms and silicon chips). Thus, the big players have created a self-imposed ceiling on their success.
For instance, when ride-sharing took off globally a few years back, China tried every type of “sharing” you could imagine, from shared concrete mixers to shared phone chargers. Some of those developments shot to the top — Mobike, for example, sold for $2.7 billion last spring — but most were “fail fasts.”
What sets the U.S. apart is its focus on disrupting an industry level.
For instance, the most well-funded AI startup in the U.S. is the B2B company Nuro, which develops autonomous vehicles focused on last-mile delivery for local commerce. Iterating innovations at an industry level usually yield longer-term results.
That’s not to say China hasn’t dabbled in B2B AI applications. The Chinese government has been pouring money into things like facial recognition. Two years ago, more than 50 cities in China participated in Xue Liang, or “sharp eyes.” The plan involved processing footage from security cameras in public and private places in order to monitor people and events.
Initiatives like these raise concerns not only around privacy but also data quality and diversity. Consumers are beginning to demand more control of their personal information (and faces are, perhaps, about as personal as you can get). While data may be “the new oil,” an AI-operated facial recognition program that’s fed an endless amount of data on Chinese faces has limited applications due to the data quality not meeting the bar.
In general, enterprises at the industry level collect and generate data in a cleaner, higher-quality fashion than the consumer level might.
Additionally, AI applications still require continuous, regular feedback. An enterprise user could provide more concentrated, constructive feedback than the individual consumer.
Finally, because we don’t have full control over AI tech, unexpected issues could arise at the consumer level. In fact, they already have. A few years back, the natural language processing tools at both Facebook and Google picked up racist terms by literally learning them from consumers’ profiles. The NLP tools suggested to both Facebook and Google’s advertisers that those words should influence their campaigns.
The industry-level applications being pursued in the States have unique advantages.
Certainly, China’s consumer applications generate value and catapult AI startups to success. However, the industry-level applications being pursued in the States have unique advantages.
Much has been made of the Fourth Industrial Revolution. It will be powered by artificially intelligent B2B solutions. For example, China is entering the healthcare space with the same consumer approach, but healthcare AI applications in the U.S. center around diagnostics.
One company I’ve invested in uses computer vision and deep learning for medical imaging enhancement. Because the company was able to quickly gain approval from the U.S. Food and Drug Administration, it could demonstrate clear data about the accuracy, efficiency, and cost savings of the technology. Now, the solution can help empower the entire healthcare industry.
Another U.S. company is disrupting a different space (auto and transportation) by developing the first in-vehicle sensing AI that identifies nuanced cognitive and emotional states — through both voice and face recognition — of drivers and passengers to improve overall transportation safety.
Why regulatory red tape is actually beneficial
It’s easy to label the stricter regulatory environment in the U.S. as an impediment to AI innovation. However, having to gain approval from regulatory bodies like the FDA helps creates a healthier ecosystem because it forces companies to practice more ethical data collection methods and, ultimately, collect cleaner data.
Interestingly enough, former FDA commissioner Scott Gottlieb was, at one time, actually a VC partner at New Enterprise Associates. (And in fact, he returned to the firm after stepping down as FDA commissioner.) Gottlieb worked to make the FDA approval process, in general, more efficient.
Although it’s popular to complain that policymakers know nothing about technology, Gottlieb is evidence of a welcome shift. If the FDA and other regulatory bodies can continue on the trajectory that he laid out, the pace of innovation will increase in far more industries than healthcare.
For now, as we watch and wait for these things to happen, a general collaborative attitude will suffice. If nothing else, it’ll help “competitors” get closer to a common goal.
Image credit: attentie-attentie; unsplash