THE TOP TEN of 2017
“MIT’s 50 Smartest Companies of 2017” (July 2017) was the most-read post of the year by a sizable margin. So number one on our list is, itself, a list. The selections by the MIT editors included these companies as the top five: Nvidia, a chipmaker, was #1; SpaceX, the space exploration group, was #2; Amazon was #3; 23andMe, a genetic testing/biomedicine firm, was #4; and Alphabet (Google) was #5.
A wider view of technology’s future can be seen in the predominance of certain fields in the list. Almost one-third of the companies selected (30%) are busy developing intelligent machines—utilizing AI and robotics. Biomedicine work makes up 22%, and 20% on the MIT list are involved with connectivity, including Amazon, Alphabet, and Tencent (China’s biggest social network). It’s in these three areas that MIT has discerned the most impressive innovation, with AI (artificial intelligence) way out in front.
Three of the top ten articles are about AI, three are about new tech inventions, and the next pair concerns autos and automotive companies. That pair includes #5 “All-electric Cars—They’re Coming” (July 2017) and the surprise second most-read, which was “Tesla Solar v. Mercedes Solar” (May 2017). That story was about two automotive companies butting competitive heads over dominance in storage batteries for home solar systems.
The three AI articles included a “Cheat Sheet for Understanding Watson” (April 2017), which discussed the nuts and bolts (hardware and software) of the neural net of 90 servers that make up Watson’s body electric, as well as five myths about the remarkable computer.
Another post, “Brains vs AI the Rematch” (February 2017) was about another human/machine contest in which computers defeated human poker champions.
And the third AI piece, “Coming Soon: AI Everywhere” (August 2017) was about a wall-mounted device called the VergeSense that creates AI-empowered (managed) workspaces.
Also making it to the top ten posts were three inventions from three industry giants. “IBM’s New Deeper Bins for Big Data” (August 2017) detailed a collaboration with Sony that produced a new way to store digital data that has set a world record for recording density. Using a Sony sputtered aluminized tape and an IBM reading mechanism, the new tape cartridges hold up to 330TB of uncompressed data in slim, four-inch cases that fit in the palm of your hand.
“Google’s Jacquard: Interactive Clothing” (March 2017) with woven digital circuits also finally made it to market.
And then there were the plans for “Jeff Bezos and his Airships” (January 2017) plus suspended hangers for Amazon.com delivery drones.
The last posting on the list is also a collection of sorts. “The Winds of Change Are Blowing” (January 2017) looks at a number of Business Insider’s (BI) predictions for 2017. Some were misses, some were accurate, and some made it halfway. In the IoT (internet of things) category, we were told to, “Look for an approved test of self-driving cars on public roads without a driver behind the wheel.” That didn’t happen, but Congress did take a major step forward in November 2017—“Congress Ready to Legislate Self-driving Cars.” Hits and misses.
Given the wide range of changes in so many areas of our lives, it’s often difficult to reasonably set our expectations. So rather than speculate on where AI and our autonomous autos will be carrying us in 2018, let’s ladle in a measure of temperance before we stir any new ingredients into the pot.
In the first week of September 2017, legendary Australian roboticist Rodney Brooks published an interesting essay in his series on the Future of Robotics and Artificial Intelligence. Brooks is former director of the MIT Computer Science and Artificial Intelligence Laboratory and founder and CTO of two companies, iRobot and Rethink Robotics. His most recent essay is titled, “The Seven Deadly Sins of Predicting the Future of AI.”
Brooks begins by commenting on some of the recent hysteria regarding robotics. He dismisses the fears centering on how powerful robots will soon become, and he also mentions a story he had read in Market Watch claiming “that robots will take half of today’s jobs in 10 to 20 years.” His curt response: “The claims are ludicrous.”
Applying his own expertise in robotics, he offers seven ways of thinking that lead to these kinds of mistaken predictions. (What follows are the seven deadly sins in digest form. I would strongly recommend that you visit Brooks’s website and read the essay.)
- First, Brooks cites Amara’s Law (Roy Amara is the President of the Institute for the Future): “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” There’s often a swing from initial exuberance to abandonment for many new technologies. He uses GPS as an example. It began as a system for the precise placement of bombs in 1978, and then after a period of disappointed expectations it quietly “seeped into so many aspects of our lives that we would not just be lost if it went away, but we would be cold, hungry, and possibly dead” as it synchronizes the U.S. electrical grid, determines what seed variants will be planted in which fields, and tracks paroled criminals.
- Next, he explains Arthur C. Clarke’s law, “Any sufficiently advanced technology is indistinguishable from magic.” For example, if you were to read a short description of how quantum computers work and then thought about Brooks’ comment on the consequences of Clarke’s law, it would become clear. “If something is magic, it is hard to know the limitations it has.”
- We often tend to think in overgeneralized terms when the subject is complicated. That’s a problem when considering AI systems. Brooks points out that those systems are very narrow in what they can do, yet many miss the limitations and overgeneralize.
- To elaborate on the tendency to overgeneralize, Brooks now considers “suitcase words.” These are descriptors that carry, for some, more than they should. Suitcase words applied to AI capabilities include terms that claim machines can anticipate, classify, estimate, hear, learn, play, read, walk, write, and so on. Brooks explains that it’s only a “narrow sliver of the rich meanings that these words imply when applied to humans. Unfortunately, the use of these words suggests that there is much more there there than is there.”
- Then there are the exponential expectations. Moore’s Law says that computing power would double annually because the number of components that could fit on a microchip would double each year. Well, that lasted for 50 years, but the expectation of exponential development in other areas of computing persists—often without any good reason.
- Hollywood scenarios have conditioned their own set of paranoid expectations that you would think reasonable people could compartmentalize in containers clearly marked “fiction.” Apparently not so, and Brooks warns, “Free running imagination about shock situations are not helpful—they will never be right, or even close.”
- Finally there’s the speed of deployment. Despite our common experience with software that undergoes constant revision and silent updating in the background on our computers, “Almost all innovations in robotics and AI take far, far longer to get to be really widely deployed than people in the field and outside the field imagine.” That’s because they are not really “thinking machines,” and the “learning” they are capable of is very, very specific and narrow.
As the new year approaches, it would probably be useful to pocket a couple of these rules and warnings as developers everywhere continue to search for ways to make our machines and solutions “smarter.”