In an often jovial chat about the future of artificial intelligence (AI), presented Wednesday at the “virtual” 2021 CES, Mobileye CEO Amnon Shashua and New York Times pundit Thomas Friedman reached agreement that before “super-intelligent” computers get smarter than people, the machines must somehow be infused with some of the “common sense” values of their flesh-and-blood inventors.
A failure to find this man/machine common ground could be, said Shashua, “a catastrophe.”
It seemed no coincidence that Shashua returned several times to the term “science fiction,” as he described the potential of AI, because speculative writers from Karel Čapek to Arthur C. Clarke and Philip K. Dick have warned readers and scientists for a century about the peril of malevolent robotic minds taking charge over human destiny.
Shashua offered an example both lighthearted and ominous of a “very sophisticated software agent powered by AI” designed to make people happier. Ashua speculated that the AI program “can figure out that if you lower people’s IQ, they tend to be less worried and may become happier.”
He continued, “This is something we did not anticipate as engineers when we programmed this AI and it’s not something that would be evident quickly. It could take decades until people’s IQ got lower… Over time, people would get dumber and dumber and dumber until we understood that this AI, which had good intentions — look, what a catastrophe.”
Check out Our Live Coverage of CES 2021
Shashua is, of course, a prominent AI advocate whose company Mobileye, is a global leader in intelligent automotive sensing systems. Friedman described his adventure in a Mobileye-equipped autonomous vehicle (AV) conducted by Shashua in the winding streets of Jerusalem, “where there are no two parallel streets… hills, hairpins, donkeys, camels, Jews, Arabs, rabbis.” Friedman enthused that Mobileye’s AV navigated this maze unscathed. He went on to explain how its sensor technology required a “complex adaptive coalition” among the machine and its human constituencies.
Mobileye, said Friedman, “actually convened an ecosystem of Volkswagen, their car supplier, the rabbis who run Jerusalem and the Israeli Ministry of Transportation. All together, as an ecosystem, they developed an insurance protocol for self-driving.”
Stretching their conversation beyond autonomous driving, Shashua and Friedman explored the imminent challenges faced by technologists, consumers and government regulators as AI systems become both more pervasive and complicated.
To characterize the swift progress of artificial intelligence, Friedman offered as an analogy the difference between classic and quantum computing, noting that AI today falls in the classic category. “Classic computing is like flipping a quarter,” said Friedman. “Zero-one, heads-or-tails. If you can flip a coin a billion times on a transistor, you’ve got compute and storage. Quantum computing is more like spinning a quarter. It can be in multiple states at the same time.”
Shashua augmented this insight by noting that the first level of comprehension in artificial intelligence was pattern recognition, a fairly basic skill. A much harder challenge is language. But Shashua said this breakthrough — machines that can read, write and tell stories — is imminent.
“This is not science fiction. In the past two years, there have been leaps in language understanding,” said Shashua. “In the next couple of years, five years at the most, I can see computers understanding text, passing reading comprehension tests. High-school reading comprehension tests are very complicated, no computer can pass it. But in two years, I believe it will.”
Shashua posited a “conversation” with a computer that reads, writes and can research exhaustively about the pros and cons of taking the Covid-19 vaccines, advising him about whether to get the shot and which vaccine variety is best for him.
The next leap in sheer computing power in AI is what Shashua and Friedman referred to as “AGI” — artificial general intelligence — a technology that uses “brute force” computing on a scale so massive that it can graduate from mimicking human behavior and tactics to conceiving its own solutions without reference to human precedent. Shashua referred to this “super-intelligence” as both “miraculous” and “dangerous.”
In a more practical vein, Friedman directed the dialog to what he called the emergence of “dual use” technology, powered by AI, and posing what he called “one of the big geopolitical and political issues in the world we’re going into.”
Friedman offered a homespun illustration. “Now, in a fast, fused, deep world, where you’ve got this acceleration in software and chip technology, everything is dual use. My toaster is dual use. If my toaster is talking to my refrigerator and I can install my toaster and refrigerator in your country’s kitchen, I can listen to you. So when we start putting intelligence into everything, everything becomes dual use. We’ve seen that in a lot of the tension between America and other countries around the world. If I sell Russia or China or any other country chips or software, how do I control the use of that?”
Friedman concluded, “Suddenly, we come back to that values question. Are we going to have healthy interdependencies or unhealthy interdependencies?”
The speakers applied this question not only to relationships among nations but the relationship between humans and their increasingly intelligent — and possibly independent — machines.
Friedman said that what Shashua calls an increasingly complicated “flat world” has few comparisons for its sheer, bewildering complexity. “The complexity of the globe today is much more mirroring the complexity of Mother nature,” said Friedman, “When climate changes, which ecosystems survive? Those that are built on complex adaptive systems.”
He said, “In human societies, those communities, countries, businesses — like Mobileye did with your coalition — that build complex adaptive coalitions to manage this change will be the ones that are going to thrive in the 21st century. And we learned that from Darwin.”
Friedman concluded by citing a global coronavirus pandemic that forced cancellation of the annual in-person CES. “With Covid-19, we’re not up against another country,” he noted. “We’re up against Mother Nature. Who does Mother nature reward? Not the smartest. Not the strongest, but the most adaptive.”
The post Shashua-Friedman Chat: AI Is Both ‘Miraculous’ and ‘Dangerous’ appeared first on EETimes.