In March, the founder and CEO of Xiaomi, Lei Jun, had a few words to say about AI. A very few.
Xiaomi is counted by some at least as being almost on the same tier as tech giants Baidu and Tencent, though its own business is centered on consumer electronics, especially smartphones and smart home devices, and there, at least, some kind of strategy on AI would seem indispensable. But while the likes of Baidu and Tencent have been charging ahead and making noise about new AI projects and acquisitions for some time now, Xiaomi has been relatively reticent on the subject. So where, and how, is the company aiming to position itself in AI?
While attending China’s annual “two sessions” legislative conference last month—where industry leaders of all stripes are invited to make advisory pitches to government—Lei offered up just three broad proposals: pushing forward with a Chinese national strategy for AI; developing retail industries to stimulate new features of the physical economy; and pushing Chinese tech companies to expand abroad.
There is plenty to be said about each of those, but for the moment we’ll just focus on the first, where Lei had some more specific recommendations, as Xiaomi revealed in a press release, suggesting that the government should:
1. Craft a national plan for top-level design and special projects in AI development
2. Strengthen theoretical research into AI
3. Cultivate AI researchers and technical specialists
4. Proactively build an AI research association and innovation community
5. Promote the transition of AI into a full-fledged industry
None of this might be worth much attention but for the fact that this was the first time Lei had spoken about AI in the better part of a year, at least publicly. And notably absent here were specifics about how AI would figure into Xiaomi’s strategic planning. Still, there are a few things to be gleaned from Lei’s proposals, and where the company might be headed.
Every company, of course, has a somewhat different take on AI, but there are two factors that every business has to take into account: data, and research talent. And for the most part, the companies with a lock on major AI research talent have been publishing papers in the areas of big data, big processing (as with GPU clusters), and big applications.
Lei had previously acknowledged that Xiaomi has established an AI exploratory lab, saying that they would “launch a heavyweight AI product before long,” although with the caveat that “before long” might mean 3-5 years.
So what is Xiaomi’s advantage? For research talent, Xiaomi may need to make some hires; for algorithms and code, it can use partnerships, or even try to bank on that “research association” Lei mentioned in his proposal. But for Xiaomi, the key asset may instead be in big data.
In answering journalists’ questions at the time, Lei said: “I think that in the AI wave, not everyone needs to be doing work on core algorithms, but on how to use AI in your field. The computational platform is already very strong … what’s the most valuable thing in the AI age? It’s big data. How much data you have. This is what generates results.”
Today, all companies and products harvest data from us to one extent or another, but some data points are more useful than others. For certain sales models, simply knowing when a user receives their paycheck and has money to burn is more valuable than a list of their interests and hobbies. For Xiaomi, with its phones in the hands of perhaps a few hundred million people, the data it can collect from smartphone usage isn’t necessarily that rare or precious, at least compared against what its competitors can obtain. But looking beyond smartphones, Xiaomi may have the single greatest advantage in China, thanks to the complete ecosystem of products it has put before its users:
“For instance, when Xiaomi users get up, when they go to bed, how much water they drink every day, what’s the temperature like in their room. But internet companies can also collect a lot of data that Xiaomi can’t, like what users like to buy, what movies they like to watch, what they like to eat. Data is never just 1 or 0, yes or no; there are a lot of dimensions.”
Lei’s dissatisfaction came through in his statement: even though big data has been its own field for years now, “progress has been less than what everyone was expecting.”
The success of Amazon’s Echo has surprised more than a few, leading to a rethink of the assumption that phones might be the best control for a smart home. It suggests that within certain scenarios at least, setting a device down in one place and leaving it there is important to creating a seamless experience, and underneath Echo’s AI technology (or Alexa’s, rather), some have begun to tease out the real shape that a smart home might take.
Jingdong’s own Linglong speaker has been called the “Chinese Echo,” using voice recognition technology from iFlytek, and Jingdong’s 3C has created an SDK for most of the smart products under the company’s umbrella to help expand their system.
Last year, a number of AI speaker startups touted Amazon Echo’s “skills” concept, but the breakthrough that’s needed is in the problem of coordination among smart home products, and creating a common focal point through which to access them all.
But there’s nothing very “smart” in merely imitating an Amazon Echo, because with the current limitations, domestically made smart speakers risk being little more than immovable smartphones on a tabletop.
Yet this is where Xiaomi stands out, because as part of its Mi Home product line it has designed a set of connected, sensor-enabled devices that seem to all but call out for the addition of an AI brain.
Thus, Lei’s approach might end up being different from that of other networking companies; different also from Baidu’s emphasis on creating big data and big everything; and almost certainly different from Nvidia’s hardware aggregation. And Xiaomi will probably not be trying to reinvent AI on its phones, but leave the work of basic AI research and big data generation to others. Instead, it looks likely to concentrate its AI efforts on its Mi Home products, leveraging the data that others have to offer. As Lei himself said: “I think we are more focused on using AI, and big data, towards specific applications.”