- The Quantum Network boots up
- Why Baidu can compete with Google
- Cornering a $1.5 trillion market
Over the last few weeks I have been drafting an investment report on the impending impact of autonomous machines.
This report pulls together everything that I’ve been talking about in my briefings — Big Data, the Cloud, AI, the Second Economy — I look forward to sending it on once my research is complete.
In drafting the report, and speaking to contacts and scientists in these fields, I’ve identified a number of critical misconceptions about technology…
Many investors recognise that power in business and industry now lies with those who can develop mathematical algorithms to harness huge volumes of streaming data.
They have read about AlphaGo and Alexa and realise that disruptive technology is on its way…
But few can explain exactly why these technologies are going parabolic.
And even fewer appreciate the challenges to scaling up these technologies. Take what has happened in China over the last decade…
The Quantum Network
Future versions of algorithms like AlphaGo that can teach themselves as they go along, will become mainstream in years to come.
They will use advanced image and language understanding skills such as we’re seeing at an embryo stage in Amazon’s Alexa and in face recognition systems at airports. They will make the predictive consumer sentiment and behavioural analytics that power the Google and Facebook advertising businesses seem unrefined.
This will require a radical leap in computing power.
At Hefei University of Science and Technology, a research hub for the PLA, there are reports by university scientists of marked progress in quantum computing, which promises to be 100 million times faster than today’s supercomputers. But we need, perhaps, to take these with pinches if not cellar loads of salt.
Nevertheless, China is scheduled to launch a putatively hack-proof quantum satellite next month, part of a quantum communications network that will be in operation between Shanghai and Beijing by the end of the year, largely for military communications.
This computer power will be key to advances in machine learning.
It took the power of 50 supercomputers to enable AlphaGo to play 1.3 million simulated games of Go, and over 1,500 to perform in the competition setting.
Deep learning pioneer Geoffrey Hinton needed 16,000 computers at Google to run an algorithm that can recognise faces from 10 million random videos.
Probably 80% of the non-linear progress in machine learning is down to the availability of virtually unlimited computer power.
In fact, scientists had cracked the theories and maths of machine learning and neural networks decades ago, but they lacked the requisite computer power until a decade ago.
Why Baidu can compete with Google
China already has several Tiahne 4 supercomputers — currently on most supercomputer metrics the world’s most powerful supercomputers — at work on weapons development and genomic projects.
And last year, Baidu built its Minwa supercomputer. The company reported that Minwa correctly identified 95.42% images in a 1 million picture base, a near 5% better hit rate than had been achieved by Google in such an exercise.
Baidu’s AI drive is serious and world class.
It is spearheaded from its $350 million R&D hub in Silicon Valley under chief scientist Andy Ng — an ex Google ‘rockstar’ and reckoned by the cognoscenti to be among the top half dozen scientists in the field.
This is another major challenge to scaling up AI — the scarcity of scientists.
Andy Ng, a disciple of Hinton and Facebook’s own Yann LeCun, helped launch the Google Brain project but was snatched up by Baidu to lead a team 200 deep learning researchers and computers systems engineers, all focused on developing algorithms that capable of unsupervised learning.
How will all this learning by used?
Well these efforts are not just in aid of Baidu’s core search business.
Like Google, Baidu is in earnest about dominating the nascent robot or autonomous vehicle business which Google famously pioneered and still leads.
Last summer, Baidu ran a converted, fully self-drive BMW3 Series on an 18.6 mile journey through Beijing, successfully performing tasks like u-turns, changing lanes, passing cars and merging on and off highways.
It was as impressive as anything done by Google or, for instance, by the Tier-1 automotive supplier Delphi.
This made possible by Baidu’s AutoBrain software, which is capable of highly automated mapping, advanced positioning, detection and smart decision making capabilities fed by a Velodyne LIDAR roof-top radar sensor — the kind used by Google.
Baidu is working with electric vehicle maker BYD, in which Warren Buffett has a hefty investment, on a range of vehicles, including buses for the US market.
It’s still an open question about when and on what scale fully self-drive production models will first appear with competing dates being bandied about by Tesla, Ford and Baidu itself, let alone when they’ll become mainstream.
But it’s clear enough that we’ve already entered the era of advanced drive assist vehicles, and we’re talking anything from dumper trucks and urban transit vehicles and from short hop robot taxis to luxury sedans.
BATS are building an algorithmic economy
Robin Li, the Baidu founder/CEO often remarks that as and when his domestic search business flat lines–and he’s very much confined to a maturing Chinese market in that–automotive could become a huge new growth business for Baidu.
He talks of an addressable market in China itself of $1.5 trillion a year down the line for vehicles and auto services of all kinds.
Indeed, like the other BATs — Alibaba and Tencent — he’s jumping into adjacent businesses such as banking and insurance. That deserves an entire briefing in itself.
At last December’s World Internet Forum in Wuzhen, Xi Jinping made a point of being seen spending time at the Baidu stand where the BMW3 Series autonomous vehicle was on display.
Global success in the global motor industry is critical to China’s economic future and its prestige.
A clear signal that in the Xi Jinping controlled Party-State, Baidu will get all the support, financial and otherwise it needs to make a good fist of becoming a dominant force in automotive.
That goes for Alibaba, Tencent and whatever strategic sectors these Tech Titans target next. Woe betide any of them, though, who fall out with Xi Jinping who seems to be operating a quasi Maoist regime.
Are you investing in China?
China figures heavily in my forthcoming report. And with good reason. I believe that the BATs will be very competitive with the likes of Google, Amazon and Facebook, and it’s important to understand the visions of these Chinese tech titans.
Each is pursuing aggressive plans to establish a platform in a new emerging economy…and the investment opportunities go beyond cars, banking and insurance.
As investors, I think it’s also important to get a handle on how the likes of Xi Jinping will seek to intervene, especially since many of these industries cross strategic interests.
Are you investing in China yourself?
I’d certainly like to gauge your interest in this area… Chinese technology seems to be an under-appreciated story.