If you’re worried about robots replacing humans, you can relax. It’s going to take a while, and there’s a fair chance it will never happen at all.
This is not just our view, but the view of Rodney Brooks, one of the world’s leading roboticists:
“We don’t have anything anywhere near as good as an insect, so I’m not afraid of superintelligence showing up anytime soon.”
Rodney Brooks knows what he is talking about. As a brilliant robotics researcher at MIT, he co-founded the iRobot corporation 28 years ago in 1990.
Brooks and the iRobot team went through 14 failed business models over the course of 12 years. But they finally hit the jackpot with not one, but two successful business models in 2002: Military robots for defusing roadside bombs and searching caves in Afghanistan, and the now-famous Roomba vacuum cleaner.
The Roomba is now the most commercially successful robot of all time. The iRobot corporation, traded under symbol IRBT, has a $2.3 billion market cap as of this writing.
In 2008 Brooks co-founded another company, Rethink Robotics, to focus on “cobots,” shorthand for “collaborative robots” which can safely work alongside humans.
The most famous Rethink Robotics product, a cobot named Baxter, had expressive eyes and eyebrows that gave it a sense of personality.
Unfortunately, Rethink Robotics didn’t follow in the footsteps of iRobot. To the surprise of the robotics community, the company shut its doors in October 2018.
It turns out that building robots with human-like capability is really, really hard. There are many things humans can do easily that are almost impossible for robots to replicate. Rethink Robotics was on a promising track, but the economics just weren’t there.
This is an example of “Moravec’s Paradox,” a principle first described by robotics researcher Hans Moravec and other pioneers (including Rodney Brooks) in the 1980s.
Here is the paradox in Moravec’s own words:
“It is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.”
Before robots can take over the world, they will need to figure out door knobs. (That still hasn’t happened yet.)
Or to put it another way: It is easier to train a machine to beat Russian chess grandmaster and former world chess champion Garry Kasparov in chess than it is to have that machine move pieces around on a wooden chessboard.
Because Moravec’s Paradox is so challenging, many robotics enthusiasts think the Turing Test, which analyzes a machine’s ability to replicate human conversation, is the wrong way to go. They argue a better challenge would be “the coffee test,” proposed by Steve Wozniak of Apple fame.
To pass the coffee test, a robot would have to enter a home it has never seen before, make its way to the kitchen, and then successfully prepare a cup of coffee.
For a person, this task would be trivial. For a modern robot it is beyond impossible.
The robot would need a generalized sense of what kitchens look like; it would need to navigate potential steps and stairs; it would need a conceptual sense of what “coffee” is; it would need to be prepared to find anything from a Keurig machine to a standard coffee maker to a water pan or a French press; it would need facility with drawers, buttons, knobs, and shelves in any combination; and it would then need an elaborate series of improvised movements to make the actual coffee.
That sequence is so far off it might as well be decades. There are countless situations like this. Something as simple as putting together a piece of mail-order furniture requires hundreds of specialized movements and on-the-fly judgments, even with high quality instructions.
Why does Moravec’s paradox hold true?
In part, because robots are so bad at this stuff — they are not within a thousand miles of true conceptual “intelligence” — and in part because human beings are wildly skilled, from birth onward, in ways that researchers are still only just beginning to understand.
When Brooks said, “We don’t have anything near as good as an insect,” he meant it literally. He spent much of his time at MIT working on insect-like robot designs to draw inspiration from nature. There is still nothing close.
If you imagine a continuum of intelligence, with, say, humans at one end and insects on the other, artificial intelligence is nowhere on that spectrum. Modern-day AI is not as smart as a bird. It is not even as smart as a bee. There is no conceptual intelligence.
It’s true that AI machines now dominate at games like Chess and have mastered video games like Pong. But what this shows is that AI in 2019 is the equivalent of a nuclear-powered calculator. It can run billions of calculations per second and crunch vast quantities of numbers faster than a human can even blink.
But that is not thinking or anything close to it. It is possible to do calculations with an abacus, a wooden tool dating to the 14th century — but nobody would ever suggest an abacus is alive or perceptive or conscious. Even today’s most impressive AI programs are little more than a turbocharged abacus (or billions of them strung together).
At the same time, humans are far more impressive than many had realized.
This is being discovered through attempts to replicate even the simple things humans can do, which turn out to be a mind-boggling challenge to even approach.
The human brain has roughly 86 billion neurons, which in turn can form approximately 150 trillion connections, in myriad different ways; it is the elegance of this design, fully integrated with the human body housing the brain, that neuroscience hasn’t even scratched the surface of.
Moravec’s Paradox — the fact that AI can parse oceans of data but can’t beat a bee or a bird, let alone a human infant — gives further insight into the future role of AI when it comes to investing.
The ability to calculate in smart ways, to spot subtle patterns with number crunching and apply algorithms to vast quantities of data is a hugely valuable thing. Machines will continue to get better and better at this, with increasingly useful outputs as a result when it comes to giant data sets (like the investable universe of stocks).
But human investors, with their ability to be creative and harness emotions and draw from unique experiences, will still have advantages the machines will always lack.
As the machines get better and the algorithms become more powerful, investing software will get better and more powerful, too. But this is something investors should look forward to rather than fear.
The big opportunity ahead involves elevating human potential rather than bypassing it. That is what great investment software is designed to do. And that’s why we see the TradeSmith mission — empowering individual investors — becoming ever more relevant in the years ahead.
Richard Smith, Ph.D.
CEO and Founder of TradeSmith