ARCHITECHT Daily: China vs. America is an AI red herring

The New York Times published a provocative article on Friday, asking in the headline "Is China outsma
ARCHITECHT
ARCHITECHT Daily: China vs. America is an AI red herring
By ARCHITECHT • Issue #86
The New York Times published a provocative article on Friday, asking in the headline “Is China outsmarting America in A.I.?”. But depending on how you define “outsmarting,” and the context in which the question is asked, the answer might not even matter much.
The answer might matter very much in terms of geopolitics and national security. Just like with supercomputing, quantum computing and other areas of deep computer science research, having better capabilities in artificial intelligence can arguably lead to an edge in areas like military, energy and climate science that can shift the world-power balance.
But if we’re talking about consumer or enterprise AI, then comparing China and the United States is kind of like comparing apples and oranges. The NYT article, for example, includes a tweet from Andrew Ng (then at Baidu) noting that Baidu achieved better-than-human-level recognition of Chinese speech a year before Microsoft claimed the same feat for English.
If this were a competition between countries (not just companies), that might seem like a loss for the United States. Except, as Ng explained to me in an ARCHITECHT Show podcast interview earlier this year, China excels in certain aspects of AI in part because there’s a confluence of cultural, technological, economic and geographic market forces that make those aspects a higher priority in China. 
Here’s an edited version of that part of my discussion with Ng:
“Because typing Pinyin, typing Chinese, on a cellphone keyboard is even more painful than typing English … I feel like China has invested, maybe even more aggressively, in advanced keyboard designs and speech recognition as an alternative for users to input information.
… Because China is a mobile-first society … today, in China, consumers are used to making very significant financial transactions [on their cellphones]. Like, you can apply for an educational loan on your cellphone, and we’ll send you a whole bunch of money … . Because of that, Baidu, and other companies, we have a strong interest in verifying who you say you are before we send you all this money.
And so that, in turn, has driven a whole lot of progress in face-recognition technology as a means to validate that you are who you say you are before we send you a bunch of money.”
“In China, there’s a very strong desire to consume non-Chinese-language content. So that, in turn, drove a lot of innovation in machine translation. For example, recently in the U.S., there was a lot of PR about … neural machine translation … . What a lot of people in the U.S. don’t know is that that was already landed, shipped to product about a year before in China.”
In the latest ARCHITECHT Show podcast, veteran investor and entrepreneur Bradford Cross said that China currently is more innovative than the United States when it comes to consumer AI. However, he also suggested that China is doing its things in consumer AI for its own reasons that don’t necessarily have a lot to do with what should be happening in the United States.
Here’s an excerpt of that part of the discussion:
“Nowadays, what’s happening is that China is driving the agenda for consumer innovation. China and Asia and also the developing world, in general, I would argue, is driving consumer innovation. And the developed world, western Europe and the U.S., in particular, are just way behind.
“And so what we’re doing is we’re starting to see the Silicon Valley folks just trying to copy China, as opposed to … the other way around. And so you actually get people trying to look at all this chatting behavior in China and people ordering pizzas via chat and all this kind of stuff and trying to replicate it here, but it’s not really a very good motivation. …
“It’s like, let’s name a trend and follow the trend and try to copy it, instead of having customer empathy and actually going to sit with the people they’re trying to serve and trying to deeply understand what the unmet need is there that you’re trying to satiate.”
I take that not as fear that China will come to dominate the AI arena and thus the world, but rather that Chinese companies are serving their consumers with cool products, and U.S. companies need to do the same—only with products that U.S. consumers actually want.
Another thing worth noting here, which the the NYT article only touches on, is the fact that much academic and even commercially funded AI research is done very much in the open. And a lot of the groundbreaking work actually comes out of Canadian universities (speaking of countries hoping to become AI hubs), as well as Google’s DeepMind division that’s headquartered in London and has ties to Oxford University. Of course, Google and Baidu also regularly push the envelope, publish papers and chronicle their innovations in AI, as do lots of other corporate and university researchers from the United States, China and other countries.
It’s a complicated issue, with a lot more to unpack in terms of global markets, geopolitical tensions (which the NYT piece gets into a bit) and cultural hurdles, among other things (including some speculation that China’s AI investments are overvalued and will result in a bubble-bursting situation). But at the risk over oversimplifying it, I think the biggest factors in how concerned you are about an AI space race is whether you think AI is a zero-sum game, and how you assess who’s winning it.
I don’t think it’s a zero-sum game, and I don’t know how you’d possibly keep score even if it was (e.g., mobile apps vs. medicine vs. driverless cars vs. agriculture). There’s simply too much money and brainpower distributed around the world (including/especially in Silicon Valley) and too much openness already baked into the world of AI research, to think that one country will somehow reap outsized rewards at the expense of everyone else.

Sponsor: Cloudera
Listen to the latest ARCHITECHT Show podcast
Merlon Intelligence CEO and veteran investor on what AI companies are doing right and wrong — and why some strategies are doomed. He also explains how Merlon is bringing automation to risk compliance.
Artificial intelligence
Hopefully, these will be some of DeepMind’s “real-world” projects around health care and energy. It will take a lot of work to make them function within the confines of industry and legal regulation.
Nvidia partnering with Foxconn, Inventec, Quanta and Wistron will certainly help with cloud-scale sales of AI boxes. But how many of those deals and buyers can it really count on?
Even if we assume ARM doesn’t stand a chance when it comes to data center AI workloads, its new designs definitely could make a dent powering smartphones and other devices. 
If you talk to anybody who knows about processors, they’ll tell you that we’re just getting started in terms of performance increases on GPUs, not to mention new types of architectures. Here’s an explanation of why that is.
Speaking of advances in AI hardware, this article is perfectly timed if you consider in the context of the three previous links. There is so much research and excitement into brain-inspired chips, but they need real-world proof points—and they need to be markedly better than the existing silicon-based approaches. 
Speaking of using the brain to improve AI, here’s a good article on some advances in studying rat—and human—brains in an attempt to build better artificial neural networks.
A total of 34 have been acquired during the first quarter of 2017 alone, according to CB Insights. As I’ve said before, I hope we’re now in a time where some of these companies can stick around, flourish, and deliver some real innovation.
This is super-important as more apps and devices become available. The FDA needs to consider the impacts AI utilization now (in terms of actual capabilities and consumer expectations), but to be able to plan for the pace of innovation.
Sponsor: DigitalOcean
Cloud and infrastructure
This is the opposite of how a digital transformation is supposed to go. It’s not clear exactly what the issue was, but anytime “penny-pinching” and “outsourcing” are being thrown around in 2017, you can bet someone didn’t get the memo about not getting Ubered.
If memory serves me, Nordstrom has been an early adopter of cloud technologies before. Here’s a proof-of-concept architecture from Nordtrom engineers for implementing AWS Lambda into an e-commerce workflow.
GitHub’s performance is probably underappreciated, if you think about how much of the world’s code it now hosts. This is a good discussion of the problems it faces (including that it doesn’t have many peers to learn from) and the solutions it’s trying. As is this ARCHITECHT Show interview with GitHub’s Sam Lambert.
If you follow the Netflix OSS projects closely, this is probably worth noting. Essentially, it’s a new orchestration engine to deal with with Netflix’s huge CI/CD volume and made for distributed processing.
medium.com  •  Share
This is an old post (well, from last month), but a good one in terms of illustrating just how much infrastructure investment it takes to compete with companies like Google and Amazon.
This is a good post from Docker, analogizing securing a house to securing software. It’s good to know Docker is thinking about this; now it just needs to make sure it can deliver on it.
Media partner: GeekWire
All things data
Since big data really hit the scene a decade ago, some folks have argued that data-derived patterns and correlations will less the importance of traditional scientific research. Here’s a good explanation of why that’s not the case.
Listen the the ARCHITECHT Show podcast. New episodes every Thursday!
Did you enjoy this issue?
ARCHITECHT
The most interesting news, analysis, blog posts and research in cloud computing, artificial intelligence and software engineering. Delivered daily to your inbox. Curated by Derrick Harris. Check out the Architecht site at https://architecht.io
Carefully curated by ARCHITECHT with Revue. If you were forwarded this newsletter and you like it, you can subscribe here. If you don't want these updates anymore, please unsubscribe here.