In some ways, this is an old topic. At least as far back as the 1960s, many people have been concerned that computers, robots, and other technology manifestations have potential to do more harm than good. It’s now 50 years since the cutting-edge machine HAL was graphically and effectively portrayed in 2001: A Space Odyssey, killing a crew member to follow its highest-priority directive of mission success, and 34 since The Terminator reinforced the dangers of what one of its characters accurately called “autonomous goal-seeking programs.” Since then it has crept into the mainstream, with products such as Amazon Echo Show and Google Home, taking over small household tasks for millions, constituting large jumps in the past year. Sixty-five-million-people Great Britain has a chance, per Jeremy Kahn in the October 14th Bloomberg.com, to add the equivalent of $837 billion to its economy with it over the next 17 years, and the potential in the United States is far higher.
Even more than with driverless vehicle technology, artificial intelligence has attracted efforts to regulate and limit it. As Andrew Burt put it in “Leave A.I. Alone” (The New York Times, January 4th), “December was a big month for advocates of regulating artificial intelligence,” with local and national bills setting the stage for its control. Yet such governing has not actually happened. The best sources now on how we should deal with it are the commentators. Here are two.
The March 31st Economist titled a 12-page “special report” “GrAIt Expectations,” which started with the observation that “artificial intelligence is spreading beyond the technology sector, with big consequences for companies, workers and consumers.” It touched on data mining, a perfect application for this technology as while it seems intense it is truly only computational, but, in only the second body-text paragraph, jumped the rails by naming a consulting company, Accenture, using it to “pick the best candidates,” something no machine, or human for that matter, can consistently do. I liked better the section’s supply-chain-progress heading “In algorithms we trust,” which is still what artificial intelligence is all about, even if modern-day computing power can do such things as determine optimal multi-stop routes, for which, with 25 locations to visit, there are 15 septillion possibilities, and replace human customer service representatives with automata in many ways better. The set of articles touched on a rapidly brewing artificial-intelligence controversy, or what we will do with information such as identifying sexual orientation, detecting unusual opinions, and potentially determining that members of certain groups may be, in general, less suited for specific employment or financial treatment. It was relatively easy for medical scientists to disregard Nazi-experiment findings, since they were less valuable, but if the latest and most powerful data mining resource were to “determine,” for example, that even when controlling for income, family background, credentials, and every other variable it can find, blacks are less successful at engineering jobs, we would have hard decisions to make.
The second piece is Tad Friend’s May 8th New Yorker “How Frightened Should We Be of A.I.?”. In this remarkably stunning and comprehensive piece, Friend started with the difference between “artificial narrow intelligence,” which harmlessly powers everything from Roombas to refrigerators, and “artificial general intelligence,” the potential 2001 or Terminator-style version, with prospects alarming even to the likes of Elon Musk, Stephen Hawking, and Alan Turing. The strictly algorithmic nature of artificial narrow intelligence, limited or not, has shown that intuition, long believed to be necessary for success in Go, is at least sometimes computational and therefore within the range of computers, including the one that beat a major champion at that game two years ago.
That intuition finding puts one thing into doubt. That is the expectation that many tasks will always require live people. Friend cited computer scientist Larry Tesler as saying that “human intelligence “is whatever machines haven’t done yet.”” As an atheist might say that religion as commonly practiced fills in only current gaps, that with today’s knowledge people no longer think that God moves planets, it could be that only our failure to understand how to reduce all human abilities to if-A-then-B thinking is stopping us from seeing that artificial intelligence can someday handle anything. Indeed, computers are already, per Friend’s citations and examples, passing the Turing test by writing in the style of petulant 13-year-olds and otherwise pretending to vary from the linear sequences we expect of them.
Three classic philosophic issues come forth as well in Friend’s article. The presence or absence of free will, or people choosing their actions themselves, may be solved by further artificial intelligence achievements. The same is true of the differences between feelings and logical thoughts. The question of whether we would want to make our planet “into a highly enriched zoo environment that’s really fun for humans to live in” may thus force itself on us. There is much more here, and I recommend anyone with interest in these topics to read it – it is at https://www.newyorker.com/magazine/2018/05/14/how-frightened-should-we-be-of-ai .
After a one-week break for the latest employment situation, I will continue this topic on June 8th with artificial-intelligence-related observations for employers, employees, and the rest of us.