China and AI

In 2010, nearly every new billion dollar start up was created in the US or Europe. But, a graph showing the rise of billion dollar start-ups till then makes one thing clear — China is clearly in second place. And, as of today, one in every 3 billion dollar start-ups is created in China.

So, what has happened in China since 2010?

As many of you know, I like approaching the future with simple mental models. And, one mental model I use from time to time is waves of foundational technologies.

The interesting thing about these waves is that they apply more to the developed world than the developing world. Most developing economies didn’t really reap the benefits of PCs and the web. Instead, many jumped straight to mobile. The mobile wave is important because it made access to cutting edge technology incredibly cheap. And, China was definitely one of them.

I could spend a ton of time writing about how mobile usage patterns in China — thanks primarily to WeChat — are different from the West. But, you likely know that. I’ll just focus on 2 stats -

• Nearly 20% of Chinese internet users are mobile only versus 5% in the United States
• 68% of Chinese users make mobile payments versus 15% in the United States

As a result, China has had a very strong and thriving mobile ecosystem. And, as technology waves go, that helps a lot as these waves often build on one another. This is especially the case with AI which, as we will see, is dependent on big company might.

Artificial Intelligence in China
Recently, a Goldman Sachs report on AI in China went viral. Here are a few highlights from the report.

First, the Chinese government aims to make China a world-leader in AI by 2030.

Geoffrey Hinton, Yann LeCun and Yoshua Bengio, the fathers of the Artificial Intelligence wave, may all be from Canada. But, China has been leading the way in research on AI.

Much of this progress is from Chinese companies — specifically, Baidu, Tencent, Alibaba and Didi.

China’s AI ecosystem is second only to the US. And, here’s a crazy stat — Baidu, Alibaba and Tencent account for 42% of the venture capital in China.

The Chinese giants spend comparable amount to the American giants (in this case, Google and Microsoft). And, they have a larger percentage of their workforce working on research and development.

So, we’re left with 3 takeaways

1. The Chinese government has decided the AI investment is important. And, we all have learnt that the Chinese government’s aims are to be taken seriously.
2. The top Chinese companies are investing heavily in research on AI and the results, so far, are positive
3. Finally, the top Chinese companies — BATD — are well positioned to compete with GAFAM for global AI supremacy. In case you’re wondering, BATD refers to Baidu, Alibaba, Tencent and Didi while GAFAM refers to Google, Apple, Facebook, Amazon and Microsoft.

AI — a one trick pony?
In an interesting article on MIT’s Technology Review (in the links below), James Somers makes a compelling point about the current AI wave being all about back propogation or backprop. Let’s start by revisiting what back propogration is with an example where the AI is tasked with identifying hot dogs -

• Imagine you have millions of pictures — some with hot dogs and some without
• You take the first image and it is the picture of a piano. Each pixel is mapped to a neuron in the bottom layer of the network.
• When you first create this “neural net,” the weights between the layers of neutrons are random — random numbers that say how much “excitement” to pass along each connection.
• As the excitement spreads up the network according to the connection strengths between neurons in adjacent layers, it’ll eventually end up in that last layer, the one with the two neurons that say whether there’s a hot dog in the picture
• The final output neurons will either be a “no hot dog” or “hot dog.” But, if the output was wrong, the errors will propagate backward so the neurons learn from the error. And, backprop is a procedure for rejiggering the strength of every connection in the network so as to fix the error for a given training example.

He goes on to explain —

Backprop is remarkably simple, though it works best with huge amounts of data. That’s why big data is so important in AI — why Facebook and Google are so hungry for it, and why the Vector Institute decided to set up shop down the street from four of Canada’s largest hospitals and develop data partnerships with them.

Hinton’s breakthrough, in 1986, was to show that backpropagation could train a deep neural net, meaning one with more than two or three layers. But it took another 26 years before increasing computational power made good on the discovery. A 2012 paper by Hinton and two of his Toronto students showed that deep neural nets, trained using backpropagation, beat state-of-the-art systems in image recognition. “Deep learning” took off.

This is why it is critical for China that Baidu, Alibaba, Tencent and Didi are investing heavily in deep learning. The AI revolution is almost certainly going to be dominated by the large tech companies. And, it is vital China’s heavy weights lead the charge.

Why does AI matter to China?
Vladimir Putin recently said something most folks who read this might agree with (not a normal occurrence) — “Artificial intelligence is the future, not only for Russia, but for all humankind,” said Vladimir Putin “It comes with colossal opportunities, but also threats that are difficult to predict. Whoever becomes the leader in this sphere will become the ruler of the world.”

I suspect, however, that the benefits of AI to China are not just about “world domination.” Consider 2 examples -

• China has installed 20 million cameras all over the country to create the world’s most advanced surveillance system. AI can help the government to identify people and help the police force search for criminals. Check this video on the Daily Mail’s website.

• China is also building a social credit system for collecting information about all its citizens from all institutions. Imagine this to be a “citizen report card” that collects your credit rating, your tax filings, your job performance, your social media history, and so on. And, of course, imagine all this is linked to your face that can be tracked using the above surveillance system.

So. I suspect the Chinese government is looking to also benefit from AI’s advances as it puts together the infrastructure to create the most powerful surveillance state in history.

How does what China is doing matter to us?
MIT Professor Rodney Brooks, in a great essay called “The Seven Deadly Sins of Predicting The Future of AI,” cautions us from over reacting to predictions around AI (link below). His essay makes a singular point very well — Predicting the future is really hard, especially ahead of time. So, beware hysteria around AI taking over the world and everyone losing their jobs.

Again, I’d like to fall back to a mental model. I go back to the question — “What will AI make cheap?” In a previous note, I drew this pretty ugly, but hopefully, useful graphic.

AI can’t do everything that humans do. But, specific applications of AI makes it cheap for machines to do things we do, better. For example, deep learning systems have a better audio transcription error rate than humans.

Similarly, to build on the China example, if we take face recognition as one application, we can build on its consequences in the near term. The Economist, in a great article on this subject, wrote — “the ability to record, store and analyse images of faces cheaply, quickly and on a vast scale promises one day to bring about fundamental changes to notions of privacy, fairness and trust.”

Here are a few examples from the article -

• FindFace, an app in Russia, compares snaps of strangers with pictures on VKontakte, a social network, and can identify people with a 70% accuracy rate.
• Facebook’s bank of facial images cannot be scraped by others, but the Silicon Valley giant could obtain pictures of visitors to a car showroom, say, and later use facial recognition to serve them ads for cars.
• Even if private firms are unable to join the dots between images and identity, the state often can. China’s government keeps a record of its citizens’ faces (as detailed above); photographs of half of America’s adult population are stored in databases that can be used by the FBI. Law-enforcement agencies now have a powerful weapon in their ability to track criminals, but at enormous potential cost to citizens’ privacy.
• Employers can already act on their prejudices to deny people a job. But facial recognition could make such bias routine, enabling firms to filter all job applications for ethnicity and signs of intelligence and sexuality.
• For example. researchers at Stanford University have demonstrated that, when shown pictures of one gay man, and one straight man, the algorithm could attribute their sexuality correctly 81% of the time. Humans managed only 61%. In countries where homosexuality is a crime, software which promises to infer sexuality from a face is an alarming prospect.
  (Note: the researcher went on record to say this study was all about proving a point)

AI makes it cheaper and easier for even the most democratic of countries to switch on surveillance.

Why we should care about the debate around AI
When John Havens, IEEE’s Executive Director for Ethical Considerations in AI and Autonomous Systems, John was asked about the biggest challenges ahead in AI. He pointed to two things -

1. As users, we have no agency or access to their data that powers AI systems. The reason AI is the playing field of massive tech firms is because said firms have access to massive amounts of user data. This means we don’t get the chance to represent ourselves in an increasingly algorithmic world. It also means the playing field is stacked against us.
2. As countries, we follow an outdated economic model built around the concept of a GDP that is wholly focused on productivity without taking into consideration metrics such as environmental sustainability and human well being. If the incentives are all about maximizing shareholder value, replacing humans with robots is a natural next step.

It felt right to share a photo of Amazon’s warehouse workers supervising robots. That robot was likely a few human workers 20 years back. The full New York Times article shared below is an amazing read/watch (as it is full of great imagery).

People who point to the industrial revolution and say we’ll find different jobs forget that there was a long period of painful readjustment.

This was a workers union demonstration in New York City in 1914 — a hundred years into the industrial revolution. And, this was not the only such demonstration around the world back then.

I get it though. It is hard to make sense of the debate on AI. For most of us, we’re working hard at our jobs, then trying to put in a good shift at home and take care of our health along the way. Maybe, if we’re lucky, we get to have a hobby or two. On the side, we hear all this buzz about various billionaires fighting each other on the prospects of AI. Is it going to lead to humanity’s doom? Is it going to bring forth the utopia where we work on better kinds of jobs? Why should we care?

In a thought provoking essay on how to think about these futurist debates, Cathy O Neil makes a telling point (lightly edited) —

“For the average person there is no difference between the singularity as imagined by futurists and a world in which they are already consistently and secretly shunted to the “loser” side of each automated decision. For the average person, it doesn’t really matter if the decision to keep them in wage slavery is made by a super-intelligent AI or the not-so-intelligent Starbucks Scheduling System. The algorithms that already charge people with low FICO scores more for insurance, or send black people to prison for longer, or send more police to already over-policed neighborhoods, with facial recognition cameras at every corner — all of these look like old fashioned power to the person who is being judged.

Ultimately this is all about power and influence. The worst-case scenario is not a vindictive AI or Sergey Brin not getting to celebrate his two-hundredth birthday. In the worst-case scenario, e-capitalism continues to run its course with ever-enlarging tools at its disposal and not a skeptical member of the elite in sight.”

Well said.

Links for additional reading

• McKinsey report on China — on McKinsey
• Goldman Sachs report on AI in China — on, well, the internet
• One in 3 unicorns is born in China — on Quartz
• Is AI a one trick pony? — on MIT’s Technology Review
• Putin on AI — on The Verge
• Unprecedented reach of the surveilance state — on ChinaFile.com
• 7 Deadly sins predicting the future of AI — on Rodney Brooks’ blog
• Life in the age of facial recognition — on The Economist
• Stanford scientist’s Gaydar was to prove a point — on Quartz
• Amazon workers babysitting robots — on New York Times (my title but awesome read)
• 5 questions with IEEE’s John Havens (AI Ethics) — on Medium
• What the Industrial Revolution really tells us about the future of work — on ACM
• Know Thy Futurist — on Boston Review
• How Starbucks’ flexible spending breaks workers — on The Nation
• How your credit scores are used by car insurance company — on Consumer Reports
• The excellent Exponential View newsletter by Azeem Azhar (I got a bunch of these links from the newsletter over the past few weeks)
• “Notes by Ada” note on AI

This is an edition of a bi-weekly technology newsletter called Notes by Ada. If you like this and would like free weekly notes via email, please just subscribe here.