Rodney Brooks knows the difference between real technological advances and unfounded hype.
Brooks is one of the world’s most accomplished experts in robotics and artificial intelligence and co-founder of IRobot, makers of the Roomba robot vacuum cleaner. Co-Founder and Chief Technology Officer of RobustAI, which makes robots for factories and warehouses; and the former director of the Computing and Artificial Intelligence Laboratories at MIT.
When Australian-born Brooks encountered a wave of unwarranted optimism about self-driving cars in 2018 — “people were saying outrageous things like, oh, my teenage son is never going to have to learn to drive” — he took it as a personal challenge. In response, he compiled a list of predictions on autonomous vehicles, artificial intelligence, robotics and space travel and promised to update them every year by January 30, 2050, when he will be 95 if he lives.
I don’t think we are ultimately limited in our ability to build humanoid robots. But whether we already have an idea how to do it, or whether all the options that we think will work are even remotely right, that’s completely open.
– Robotics and AI expert Rodney Brooks
His aim was to “inject some reality into what I saw as irrational exuberance”.
Every prediction contained a time frame – something would have happened either by a certain date, or no earlier than a certain date, or “not in my lifetime”.
Get the latest from Michael Hiltzik
Commentary on Economics and More by a Pulitzer Prize Winner.
You may occasionally receive promotional content from the Los Angeles Times.
Brooks released its fifth annual scorecard on New Year’s Day. Most of his predictions were spot on, although this time he admitted he thought he too had allowed the hype to make him overly optimistic about some developments.
“My current belief is that overall things are going to be even slower than I thought five years ago,” he wrote earlier this year.
As a seasoned technologist, Brooks has an idea of what makes laypeople, or even experts, overly optimistic about new technologies.
People have been “trained by Moore’s Law” to expect technology to improve at an ever-increasing rate, Brooks told me.
He refers to a 1965 observation by semiconductor engineer Gordon Moore that the number of transistors that fit on a microchip is doubling roughly every two years. Moore’s observation became a proxy for the idea that computing power would improve exponentially over time.
This leads people, even experts, to underestimate how difficult it can be to achieve a chosen goal, be it self-driving cars, confident robots or life on Mars.
“They don’t understand how hard it was to get there,” he told me, “so they assume it just keeps getting better.”
One example is driverless cars, a technology with limitations that laypeople rarely realize.
Brooks has written about his experiences with Cruise, a self-driving taxi service (no one in the front seat at all) in parts of San Francisco, Phoenix and Austin, Texas.
In San Francisco, Cruise only operates between 10 p.m. and 5:30 a.m.—that is, when traffic is at its lightest—and only in limited parts of the city and in good weather.
On his three Cruise rides, Brooks found that the vehicles avoided left turns, instead taking three right turns around a block, driving painfully slowly, and once attempting to pick him up in front of a construction site that would have exposed him to oncoming traffic.
“The result is that it was a factor of two slower than any human-operated ride-hailing service,” Brooks wrote. “It may work for select regions, but it won’t compete with human-operated systems for quite a while.” It’s also “decades away from viability,” he judged. In this year’s scorecard, he predicted that “human drivers will be on our roads for decades to come.”
The annual scorecard is one of many sources Brooks relies on to mitigate the “irrational exuberance” about technology in general and AI in particular. He is a regular contributor to IEEE Spectrum, the house organ of the premier professional association of electronics engineers.
For example, in a September 2021 article titled “An Inconvenient Truth About AI,” he noted that each wave of new developments in AI has been accompanied by “breathless predictions of the end of human dominance in intelligence,” amid “a tsunami.” of promises, hype and profitable applications.”
In reality, Brooks wrote, almost every successful use of AI in the real world has had either a human “knowledge somewhere” or very little cost of failure. The Roomba, he wrote, functioned autonomously, but its worst failure might be “missing a patch of ground and not picking up a ball of dust.”
However, when IRobots were used in Afghanistan and Iraq to disable improvised explosive devices, “errors there could kill someone, so there was always a human on the loop issuing surveillance commands.”
Robots are now widely used in industry and even in the home, but their capabilities are very limited. Robot hands with true human-like dexterity haven’t evolved much in 40 years, Brooks says. This also applies to autonomous navigation in every house with all its clutter, furniture and movable objects. “What’s easy for humans is still very, very hard for robots,” he writes.
Regarding ChatGPT, the AI prose generator that has attracted undue interest from high-tech enthusiasts, along with warnings that it could usher in a new era of machine-driven plagiarism and academic forgery, Brooks urges caution.
“People make the same mistake they’ve made over and over again,” he writes in his scorecard, “of completely misjudging some new AI demo as a sign that everything in the world has changed.” It didn’t.”
ChatGPT, he writes, replicates patterns in a human prompt rather than displaying a new level of intelligence.
That’s not to say Brooks doubts the eventual creation of “truly artificial intelligences, with cognition and consciousness discernibly similar to our own,” he wrote in 2008.
He expects that “robots that will roam our homes and workplaces … will emerge gradually and symbiotically with our society,” even as “a wide array of advanced sensory devices and prosthetics” emerge to enhance and augment our own bodies: “So as our machines become more like us, we will become more like them. And I’m an optimist. I think we’ll all get along.”
Which brings us back to Brooks’ 2023 scorecard. This year, 14 of his original predictions are believed to be correct, either because they came within the timeframe he predicted or because they didn’t come before the deadline he set.
These include driverless parcel delivery services in a major US city, which he doesn’t forecast before 2023; it hasn’t happened yet. Regarding space travel and space tourism, he predicted suborbital human launch by a private company by 2018; Virgin Atlantic’s deadline for such a flight is 12-13-2018.
He surmised that space flights with a handful of paying customers wouldn’t happen until 2020; regular flights at a rate of more than once a week not before 2022 (but maybe by 2026); and the transportation of two paying customers around the moon no earlier than 2020.
All of these deadlines have passed, so the predictions are accurate. Only three flights with paying customers took place in 2022, showing that “there’s a long way to go to get to weekly flights,” Brooks notes.
Brooks is consistently skeptical of the projections of our most-cited tech entrepreneur, Elon Musk, who Brooks says has “a pattern of overly optimistic time-frame predictions.”
A moon orbit for paying customers in Musk’s SpaceX Falcon Heavy capsule does not appear to be possible before 2024, Brooks observes. The landing of cargo on Mars for later human use, which Musk once predicted for 2022, looks like it won’t happen until 2026, and even that date is “way too optimistic.”
Musk still hasn’t delivered on his 2019 promise that Tesla would have 1 million robo-cabs on the road by 2020 — a fleet of autonomous cars summoned via an Uber-like Tesla app. “I believe the actual number is still a solid zero,” Brooks wrote.
As for Musk’s dream of running regular service between two cities on his Hyperloop underground transportation system, Brooks puts it in the “not in my life” box.
Several Brooks predictions remain open, including some involving the electric vehicle market. In his original forecast, he projected that EVs would not reach 30% of US auto sales until 2027 and 100% until 2038.
The growth rate in electric vehicle sales accelerated in 2022 — rising 68% in the third quarter from the same quarter last year. If this growth rate continues, electric vehicles would account for 28% of new vehicle sales in 2025.
That assumes that the forces driving EV adoption persist. However, the headwind should not be underestimated. Electric vehicle sales may have skyrocketed in 2021 and last year due to the huge increase in gasoline prices, but that inflation trend has now disappeared. Battery factories could take longer to start up than expected, leading to shortages of these vital components and driving up EV prices.
“There’s clearly something afoot,” Brooks writes, although “it’s not clear yet” if the U.S. will reach 30% EV market share by 2027.
Brooks doesn’t want to stifle human endeavors to build robots, AI systems, or space exploration.
“I’m a technologist,” he told me. “I build robots – that’s what I’ve done with my life – and I’ve always been a space fanatic. But I don’t think it does people well to be so ‘overly optimistic’ that they ignore the hard issues standing in the way of progress.
“I don’t think we’re ultimately limited in our ability to build humanoid robots,” he says. “But whether we now have an idea of how to do it, or whether all the options that we think will work are even remotely right, that’s completely open.”
He compares the dream to that of medieval alchemists trying to turn lead into gold. “You can do it now with a particle accelerator to change atomic structures, but back then they didn’t even know atomic structure existed. We may be like that on a human level, but we have absolutely no idea how it works.