The threat of artificial intelligence
by Christopher Joye
While we profit from Homo sapien's current living standard and longevity peaks after hundreds and thousands of years of evolution—we're the most advanced and dominant species to populate planet earth—many esteemed scientific experts believe the single greatest threat to our existence is the so-called "Singularity".
This is the event predicted to arrive between 2020 and 2050 when the exponential increase in computing power—explained by "Moore's law" that forecasts that the number of transistors on integrated circuits doubles roughly every two years—could enable "narrow" artificial intelligence (AI) to evolve to, and critically surpass, "strong" or "generalised" AI that is equivalent to human functionality.
At this game-changing juncture non-biological intelligence would pass a "Turing Test", demonstrating through natural language conversations that it is indistinguishable from a human. It is AI that would exhibit self-awareness and a desire to avoid "expiry". With access to global computing networks, knowledge and resources via the Internet, human-level AI could embark on both astonishingly fast "recursive self-improvement" and self-replication to safe-guard its own sustenance—that is, it would continuously reprogram itself to optimise its powers.
This would be akin to accelerated human evolution and reproduction: the difference being that the AI can instantaneously propagate itself with trivial cost around the world just as digital (and, less effectively, biological) viruses do. And with near-infinite memory and perfect recall it could self-improve much more rapidly than humans can. The Singularity is thus the point at which AI materially exceeds human capabilities and control. And since this new super-intelligence would, by definition, be beyond our comprehension, it is impossible to predict what might unfold.
The mathematician Vernor Vinge first outlined the concept for a NASA-sponsored symposium in his 1993 essay, "The Coming Technological Singularity". This prophesied that "within thirty years, we will have the technological means to create superhuman intelligence" with the downside that "shortly after, the human era will be ended".
You benefit from "narrow" or "weak" AI every day unwittingly. Google and Facebook use AI in their search, speech recognition, translation, newsfeed and image identification tools, and are two of the biggest commercial backers of AI research. iPhone owners use Apple's intelligent personal assistant, called "Siri" for "Speech Interpretation and Recognition Interface", to get answers to questions like where's the closest cafe or what's the distance to my next meeting.
Others examples include: autopilot software in aircraft; the "algorithmic" trading programs that dominate sharemarket volumes today, provide automated execution (displacing traditional stock brokers) and implement the quantitative trading strategies that caused the 2010 "flash crash"; the computer-generated voice you hear when you call a bank, which outlines options, answers questions, and guides your inquiries; the parallel parking systems found in cars like the Jeep Cherokee; the new forms of "robo-advice" that produce detailed financial plans based on your risk preferences; and autonomous "drones", like Australia's Global Hawk spy plane that can fly itself for as long as 30 hours to altitudes above 60,000 feet or the US Navy's X-47B unmanned fighter jet that can land and take-off from a moving aircraft carrier, and perform attack missions without active human flight control. In the future I think Australia's two new mini aircraft carriers, HMAS Canberra and HMAS Adelaide, will become "mother-ships" for vast numbers of drone fighters.
Creating strong AI that can emulate our brains is a major focus for academic, commercial and military researchers. So how worried should we be about the ensuing Singularity? Billionaire technology entrepreneur and Tesla founder, Elon Musk, told an MIT conference "we should be very careful about AI". "If I had to guess at what our biggest existential threat is, it's probably that", he said, arguing for greater regulation of the sector.
"With artificial intelligence we are summoning the demon," Musk colourfully claimed. "In all those stories where there's the guy with the pentagram and the holy water, it's like – yeah, he's sure he can control the demon. [But it] doesn't work out."
Celebrated Cambridge physicist Stephen Hawking, who suffers from motor neurone disease and harnesses AI to speak, is no less dire. In December 2014 he told the BBC that "the development of full AI could spell the end of the human race." "It would take off on its own, and re-design itself at an ever increasing rate," he said. "Humans, who are limited by slow biological evolution, couldn't compete, and would be superseded."
In May 2014 Hawking co-authored an open letter with three other leading academics, including a Nobel prize winner, warning that dismissing the risks of intelligent machines portrayed in movies like Transcendence, which I have previously analysed here, as science fiction "would be a mistake, and potentially our worst mistake in history".
While the co-authors characterised the advent of strong AI as "the biggest event in human history", they concurrently cautioned it "might also be the last, unless we learn how to avoid the risks". "One can imagine such technology outsmarting financial markets, out-inventing human researchers, out-manipulating human leaders, and developing weapons we cannot even understand," they said. "Whereas the short-term impact of AI depends on who controls it, the long-term impact depends on whether it can be controlled at all."
Other eminent minds ringing the same alarm bells include: Bill Joy, co-founder and former chief scientist at Sun Microsystems, in his controversial Wired magazine feature entitled, "Why the future doesn't need us"; Ray Kurzweil, the Google exec and acclaimed AI developer that invented speech recognition software, who thinks "the Singularity is near"; and billionaire Microsoft founder Bill Gates, who last year declared "I am in the camp that is concerned about super intelligence".
Gates believes initially the "machines will do a lot of jobs for us and not be super intelligent", which will be "positive if we manage it well". "A few decades after that, though, the intelligence [will be] strong enough to be a concern," he predicts. "I agree with Elon Musk and some others on this and don't understand why some people are not concerned," Gates says.
Parallel innovations, like ultra-fast quantum computers (expected within our lifetimes), could accentuate the AI threat. Another is nanotechnology, which the new book, The Rise of the Robots, explains involves building machines ("nanobots") at the molecular level that can rearrange atoms into almost any physical object from cheap resources. This would replicate the naturally occurring "ribosome" molecular factory contained within cells, which reads information encoded in DNA and then assembles the "building blocks of all biological life".
In Wired for War, Dr Peter Singer, a cyber security expert who has advised the CIA and the US Defense Department, quotes Australian researcher Dr Hugo de Garis who says that within a single generation it will "very probably be possible to store a single bit of information on a single atom". This would allow a disc-sized object to hold a trillion, trillion bits of information, dwarfing the 23 million bits of information in the brain's genome and possibly bequeathing a machine with sufficient power and storage capacity to rival our capabilities.
In 2005 professor de Garis, who researched "genetic algorithms" that mimic natural selection to evolve "neural networks" that solve problems like a brain, published a paper warning that by the end of the 21st century there could be a war between humanity and intelligent machines. And this was no crank: he went on to run the Artificial Brain Lab at China's Xiamen University!
Over the years this columnist has encouraged investors to consider unlikely events, including geopolitical risks (borne out in Russia's invasion of the Ukraine, Middle Eastern conflicts, North Korea's nuclear detonations, and China's ructions) and public and private cyber-attacks, which are now commonplace.
I've proposed solutions, like avoiding mindless diversification into regions such as Asia (China, Japan and Korea specifically), investing in cyber-security companies, and imploring governments to commit real effort into analysing and forecasting the historical frequency and future probability of conflicts occurring.
Like the opportunities afforded by nuclear energy, autonomous robotics could positively contribute to our lives. Yet as with the atomic bomb, there is a dark side that we need to monitor: the risk that human evolution engineers a new species of intelligent artificial life that ironically supercedes us.