- A new era of Intelligent Machines
- Why Google wants to radically scale up Search
- Up to 40% of companies in mature industries at risk
I believe we are transitioning towards a society that is heavily dependent on intelligent machines.
And there are immediate consequences for investors — not least the fact that many companies in mature industries face being eviscerated in the next few years.
Take the breakthrough that we saw last week with Google’s AI machine…
30 Million Moves to Beat a Human
When Google’s AlphaGo machine learning algorithm beat the world ‘Go’ champion 4-1 last week even hard-nosed computer scientists went slack jawed.
It was something that until very recently had been thought to be a decade away.
AlphaGo had learnt the ancient Chinese game by playing tens of millions of games against human players and itself, relying on a “convolutional neural network” to mimic the way neurons talk to each other in the brain.
Like an infant brain, AlphaGo was programmed to recognise images and learn under basic instruction. Left to its own devices, the algorithm reinforced its learning by repeatedly playing against itself, achieving skills in pattern recognition and calculation that were previously the domain of human intelligence.
This was an algorithm that, to a large extent, had taught itself to play a game that has confounded us for centuries.
AlphaGo has raised the profile of Artificial Intelligence, but in all the excitement many people have missed an even bigger story.
50 Billion Devices Online by 2020
Google, the parent company of AlphaGo, has always seen itself as an AI company.
Larry Page recognised early on that the Search engine functioned like AI — an intelligence that constantly reinforces itself with each ‘search’ and nugget of data.
The real story is how Google has been scaling up its machine learning efforts for some time.
The popularity of smartphones has created a vast new swarm of devices to draw data from. There were 14.8 billion devices connected to the internet in February last year, according to Cisco, and that number is set to explode with the advent of the Industrial Internet of Things.
This social network of giant locomotives, jet engines, wind turbines and robots will throw off a volume of data that will dwarf what people generate from the conventional Internet. Cisco estimates that there will be 50 billion devices online by 2020.
Economist Brian Arthur calls this great hive of sensors and devices “The Second Economy” — a vast, automatic and mutating global web of machines in constant communication with each other. And free of much external control, the Second Economy will soon transact a huge proportion of the world’s business with little or no human intervention.
Let Machines Teach Themselves
The Second Economy demands a new kind of machine intelligence.
In fact, just last month the man who drove Google Search, Amit Singhal, announced that he is retiring, to be replaced by John Giannandrea, the man has been overseeing Google’s work in deep neural networks.
Under Singhal, Google Search resisted the introduction of machine learning, preferring algorithms that rely on a strict set of rules, dictated by humans. His concern was that we don’t really know why neural networks work. You can feed them with pictures of a board game like Go and it can learn to play the game, but we don’t know exactly why. That makes these algorithms difficult to control and tweak.
Giannandrea believes in allowing the machines to learn by themselves. Breakthroughs in neuroscience and the availability of huge volumes of data on top have enabled a step change in machine learning. Hand-crafted programs where human programmers distill information from experts into specific rules and heuristics are no longer the way to do machine intelligence.
The AI Race
Tech Titans such as Alphabet/Google, Amazon and Facebook are currently far ahead of the rest of the online platform pack. They have such footholds in their respective markets that they are virtually impossible to dislodge, at least in the medium term.
The application of machine learning will vastly enhance their businesses. By applying its machine learning algorithms to vast streams of data, Google can start to match humans in a variety of tasks.
They are making impressive advances in cognitive computing – the simulation of human thought processes by a computer – identifying photos, recognising commands spoken into our smartphones, anticipating our needs.
But they also need to radically scale up Search in order to adapt to an economy increasingly controlled by machines.
Google’s competitors are hard at work on their own AI projects.
IBM, Facebook, Apple, Microsoft, Amazon, Samsung, Tencent, Baidu and others are serious AI cheerleaders, sporting large user bases, and the necessary resources to develop and deploy.
Each operates on the economic logic of ‘increasing returns to scale’: the more people who use their systems, the smarter these systems become and the smarter they become and the more people use them…
The application of machine learning will open up new opportunities in other areas such as social robotics, personal medicine and healthcare.
- IBM famously distilled information from Chess grand masters and Jeopardy champions into specific rules and heuristics to enable Deep Blue and Watson to beat all-comers. Watson is currently finding work in F1 pits, hospitals and in addressing anti-fracking attacks on Twitter. It is already establishing itself as AI for companies in the Second Economy to tap into, even if it still relies on rules set by humans.
- And, in a move timed to exploit the national obsession with the recent Google ‘AlphaGo shock’, the South Korean government has announced plans to spend $863 million over the next five years to turn Korea into a leading AI society. This will include the creation of a public-private research centre with support from Samsung, LG Electronics and Hyundai Motor.
- The team that developed Apple’s Siri recently broke away to form Viv Labs with backing from among others Iconiq Capital > and Li Ka-Shing’s Horizon Ventures. An impressive roster. Viv Labs is developing a super-Siri that is ”taught by the world, knows more than it is taught and learns new things every day” and can engage with any appropriate device. Both Google and Facebook are reported to have tried and failed to acquire it.
The winners in this new AI race will be those businesses with the most advanced proprietary algorithms, able to work on huge and fluid datasets that are generated by humans AND machines.
The Risks to Your Portfolio
While there are clear opportunities for growth, there will be grave risks too.
In these times of rapid, non-linear change the average age of S&P 500 companies is 12 years and falling vs 40 years when the index was formed 40 years ago.
I believe that within 10 years, up to 40% of companies in mature industries will be blown away in the technology storm.
A company’s relative tech prowess will chiefly determine whether it can create, keep and grow customers profitably and thus survive, let alone thrive.
This calls for new analytical tools to assess whether a company is a buy, hold or sell; tools that go way beyond CEO and CTO guidance and balance sheet forensics. To compete in this new environment, companies will need to foster an algorithmic culture (a staff that complements computer scientists with neurocomputational experts, mathematicians and data scientists), deep and wide user bases, and a fast evolving technology strategy.
There are trillions at stake and a coming socio-economic upheaval to contend with as more and more jobs are replaced by intelligent machines.
If you haven’t had a chance to read Brian Arthur’s fascinating article on The Second Economy, I highly recommend you do so. Arthur believes it will wipe out 100 million jobs and be the same size as the physical economy by 2030.
In the next briefing, I’ll reveal how Amazon is reshaping society around Artificial Intelligence and what it means for investors.