An extract from our 2017 Annual Letter
*from Steve Job’s “Lost Speech”, 1983, which foretells the rise of mobile computing and the iPad.
Anyone with grey hair (or no hair) will remember the hype and promise of AI through the 1970’s and 80’s. Recall: Hal 9000 (2001: A Space Odyssey), SkyNet (Terminator); C-3P0 (Star Wars) and Deep Thought (A Hitchhikers Guide to the Galaxy).
Such attempts to replicate human consciousness (known as Strong AI) indeed failed to deliver, leading to an ‘AI Winter’ during the 1980’s. One infamous AI fail was biblical in proportions: Matthew 26:41 “the spirit is willing but the flesh is weak” machine translated to “the vodka is strong but the meat is soft” when reversed from Russian.
However, in recent decades a quiet revolution has been taking place in Applied AI, led by engineers trying to solve practical problems.
Part I: AI in from the cold: from ‘AI Winter’ to ‘AI Wash’
Huge advances are occurring across 1000’s of applications. Think of: search engines; the genome project; space/military robotics; predictive marketing; facial recognition; medical diagnosis; fraud detection, and; natural language processing. G2 would add stock-picking to the list, naturally.
The world’s foremost players of Checkers, Chess, Jeopardy!, Go and Texas Hold’em poker are all computers.
But who drives the AI to the tournament?
Well, autonomous vehicles are fast [sic] becoming the world’s safest drivers … with Google’s Waymo self-driving car project logging two million miles with only one (1) at-fault accident – an accident rate 40x lower than new drivers.
While the full story is much more nuanced than the headlines, AI is indeed underpinning a genuine revolution – though one which is not well understood.
Die-hard skeptics and blue-sky evangelists get much of the attention, but as always the truth is somewhere in the middle. For G Squared, 6 years in the business of applying AI to systematize stock picking has taught us a few lessons from the coalface. Far from being techno-utopians; our mindset is much more engineering than evangelical.
Like the curmudgeonly old-timer, our eyes roll a little when extravagant claims are attached to AI. Moreover, as Mark Hurd, Oracle’s CEO, complained recently: “Everyone in silicon valley says they are doing AI - most of the time it’s just nonsense - but AI is real and it will change the way we work and live.” We agree on both points.
While marketing spin - you might say an ‘AI Wash’ - might help products sound sexy, outside of dating, sexiness is the enemy of results (as per last year’s letter).
To help separate signal from the noise, we address some popular misconceptions here.
Part II: AI - what is that again?
Artificial intelligence – and more broadly the fourth industrial revolution of which it is a central part – is a diffuse and bottom up process driven by disruptive solutions to real-world challenges.
It is a natural evolution of technology as human beings have been automating tasks for the past 2 million years: first physical now cognitive tasks. Along the way progress has been exponential not linear.
As a result AI is lots of little things. It’s not ‘one big thing’. There is no convergent solution or mega-mind Hal 9000 in sight. Sorry Googlers.
Indeed, what people refer to as “AI” is really a broad church of tools and techniques to solve specific problems. A broad classification as it relates to finance is as below:
Some insiders will quibble at what is included – and missing – but the main message is that AI is a toolkit not a single hammer or some kind of magic wand.
G2’s systematic approach to stock-picking applies various machine learning approaches. There is no single ‘magic algo’. We have 26 virtual analysts which compete to select stocks and 45 virtual traders who help cut losses and time trades. It’s an ensemble and layered approach.
Speaking of AI in finance: while it is talk of the town, our estimate is that less than 0.1% of the $80 trillion in global equity capital is deployed using AI. Humans (discretionary managers) still rule the roost when it comes to investment decision making.
“With the exception of a few leading hedge funds, the industry has failed to recognize machine learning's potential to drive investment decisions.”
- Pension & Investments, December 2017
Mr. Market is still very much human.
Part III: AI - smarter than humans or just less dumb?
Even the best AI system is not really ‘smarter than humans’ in a pure sense. It just makes fewer mistakes.
“Machines are not prophets; machines can solve a game or goal. They can make mistakes, but when you look at the average quality of the moves it’s fairly consistent. So consistency is what is deadly for humans, because even the best of us are not consistent.”
- Garry Kasparov, Business Insider, 2017
Kasparov is right – and he is a huge convert – when IBM’s Deep Blue defeated him 20 years ago in May 1997, he referred to it as a “$10m dollar alarm clock”.
In the face of defeat at chess, theoreticians and sceptics moved their defensive perimeter back to more complex games, including, the Queen of them all – the game of Go.
“It may be 100 years before a computer beats humans at Go … maybe even longer.”
- Piet Hut, Institute for Advanced Studies at Princeton University, 1997.
Oops. AlphaGo, in its 2016 victory against Lee Sedol, one of the most decorated Go players alive, made a decision (on its 37th move) that baffled all the experts. AlphaGo’s lead researcher placed the odds of a human making the same move at 1-in-10,000, yet it ultimately resulted in victory.
Let’s not even mention AlphaGo Zero, released in 2017, which surpassed all prior versions in 40 days by playing games against itself starting from completely random play, rather than being trained on thousands of human amateur and professional games.
Here again Kasparov is right – Zero is arguable the strongest Go player in history, because it is no longer constrained by the limits of human knowledge. Go is very complex, and to play without errors is nearly impossible.
But – and it’s a big one – AI may make less mistakes and be less dumb than humans – but that doesn’t make it infallible, magic or omnipotent.
We think fears that sentient AI and robot wars will destroy humanity are pretty far fetched. Elon Musk has been a high profile serial-offender: “Summoning the demon … Our biggest existential threat”. Come on Elon. The technological singularity is as far away as it has ever been … or.
In a sense this brings us full circle – back from the machine to the human. In all of this, no matter how developed AI applications become, it is truly humanity that is developing. The real intelligence lies in humans for creating and building, rather than in AI’s capabilities.