However you look at it, the future appears bleak. The world is under immense stress environmentally, economically and politically. It’s hard to know what to fear the most. Even our own existence is no longer certain. Threats loom from many possible directions: a giant asteroid strike, global warming, a new plague, or nanomachines going rogue and turning everything into grey goo.
Another threat is artificial intelligence. In December 2014, Stephen Hawking told the BBC that “the development of full artificial intelligence could spell the end of the human race… It would take off on its own, and redesign itself at an ever increasing rate. Humans, who are limited by slow biological evolution, couldn’t compete, and would be superseded.” Last year, he followed that up by saying that AI is likely “either the best or worst thing ever to happen to humanity”.
Advertisement
Other prominent people, including Elon Musk, Bill Gates and Steve Wozniak, have made similar predictions about the risk AI poses to humanity. Nevertheless, billions of dollars continue to be funnelled into AI research. And stunning advances are being made. In a landmark match in March, the Go master Lee Sedol lost 4-1 to the AlphaGo computer. In many other areas, from driving taxis on the ground to winning dogfights in the air, computers are starting to take over from humans.
Hawking’s fears revolve around the idea of the technological singularity. This is the point in time at which machine intelligence starts to take off, and a new more intelligent species starts to inhabit Earth. We can trace the idea of the technological singularity back to a number of different thinkers including John von Neumann, one of the founders of computing, and the science fiction author Vernor Vinge. The idea is roughly the same age as research into AI itself. In 1958, mathematician Stanisław Ulam wrote a tribute to the recently deceased von Neumann, in which he recalled: “One conversation centered on the ever accelerating progress of technology and changes in the mode of human life, which gives the appearance of approaching some essential singularity… beyond which human affairs, as we know them, could not continue” ().
“There are reasons to be fearful of computers“
More recently, the idea of a technological singularity has been popularised by Ray Kurzweil, who predicts it will happen around 2045, and Nick Bostrom, who has written a bestseller on the consequences. There are several reasons to be fearful of machines overtaking us in intelligence. Humans have become the dominant species on the planet largely because we are so intelligent. Many animals are bigger, faster or stronger than us. But we used our intelligence to invent tools, agriculture and amazing technologies like steam engines, electric motors and smartphones. These have transformed our lives and allowed us to dominate the planet.
It is therefore not surprising that machines that think – and might even think better than us – threaten to usurp us. Just as elephants, dolphins and pandas depend on our goodwill for their continued existence, our fate in turn may depend on the decisions of these superior thinking machines.
The idea of an intelligence explosion, when machines recursively improve their intelligence and thus quickly exceed human intelligence, is not a particularly wild idea. The field of computing has profited considerably from many similar exponential trends. Moore’s law predicted that the number of transistors on an integrated circuit would double every two years, and it has pretty much done so for decades. So it is not unreasonable to suppose AI will also experience exponential growth.
Like many of my colleagues working in AI, I predict we are just 30 or 40 years away from AI achieving superhuman intelligence. But there are several strong reasons why a technological singularity is improbable.
The “fast-thinking dog” argument
Silicon has a significant speed advantage over our brain’s wetware, and this advantage doubles every two years or so according to Moore’s law. But speed alone does not bring increased intelligence. Even if I can make my dog think faster, it is still unlikely to play chess. It doesn’t have the necessary mental constructs, the language and the abstractions. Steven Pinker put this argument eloquently: “Sheer processing power is not a pixie dust that magically solves all your problems.”
Intelligence is much more than thinking faster or longer about a problem than someone else. Of course, Moore’s law has helped AI. We now learn faster, and off bigger data sets. Speedier computers will certainly help us to build artificial intelligence. But, at least for humans, intelligence depends on many other things including years of experience and training. It is not at all clear that we can short circuit this in silicon simply by increasing the clock speed or adding more memory.
The anthropocentric argument
The singularity supposes human intelligence is some special point to pass, some sort of tipping point. Bostrom writes: “Human-level artificial intelligence leads quickly to greater-than-human-level artificial intelligence… The interval during which the machines and humans are roughly matched will likely be brief. Shortly thereafter, humans will be unable to compete intellectually with artificial minds.”
If there is one thing that we should have learned from the history of science, it is that we are not as special as we would like to believe. Copernicus taught us that the universe does not revolve around Earth. Darwin showed us that we are not so different from other apes. Watson, Crick and Franklin revealed that the same DNA code of life powers us and the simplest amoeba. And artificial intelligence will no doubt teach us that human intelligence is itself nothing special. There is no reason to suppose that human intelligence is a tipping point, that once passed allows for rapid increases in intelligence.
Of course, human intelligence is a special point because we are, as far as we know, unique in being able to build artefacts that amplify our intellectual abilities. We are the only creatures on the planet with sufficient intelligence to design new intelligence, and this new intelligence will not be limited by the slow process of human reproduction and evolution. But that does not bring us to the tipping point, the point of recursive self-improvement. We have no reason to suppose that human intelligence is enough to design an artificial intelligence that is sufficiently intelligent to be the starting point for a technological singularity.
Event:
Even if we have enough intelligence to design super-human artificial intelligence, the result may not be adequate to precipitate a technological singularity. Improving intelligence is far harder than just being intelligent.
The “diminishing returns” argument
The idea of a technological singularity supposes that improvements to intelligence will be by a relative constant multiplier, each generation getting some fraction better than the last. However, the performance of most of our AI systems has so far been that of diminishing returns. There are often lots of low-hanging fruit at the start, but we then run into difficulties when looking for improvements. This helps explain the overly optimistic claims made by many of the early AI researchers. An AI system may be able to improve itself an infinite number of times, but the extent to which its intelligence changes overall could be bounded. For instance, if each generation only improves by half the last change, then the system will never get beyond doubling its overall intelligence.
The “limits of intelligence” argument
There are many fundamental limits within the universe. Some are physical: you cannot accelerate past the speed of light, know both position and momentum with complete accuracy, or know when a radioactive atom will decay. Any thinking machine that we build will be limited by these physical laws. Of course, if that machine is electronic or even quantum in nature, these limits are likely to be beyond the biological and chemical limits of our human brains. Nevertheless, AI may well run into some fundamental limits. Some of these may be due to the inherent uncertainty of nature. No matter how hard we think about a problem, there may be limits to the quality of our decision-making. Even a super-human intelligence is not going to be any better than you at predicting the result of the next EuroMillions lottery.
The “computational complexity” argument
Finally, computer science already has a well-developed theory of how difficult it is to solve different problems. There are many computational problems for which even exponential improvements are not enough to help us solve them practically. A computer cannot analyse some code and know for sure whether it will ever stop – the “halting problem”. Alan Turing, the father of both computing and AI, famously proved that such a problem is not computable in general, no matter how fast or smart we make the computer analysing the code. Switching to other types of device like quantum computers will help. But these will only offer exponential improvements over classical computers, which is not enough to solve problems like Turing’s halting problem. There are hypothetical hypercomputers that might break through such computational barriers. However, whether such devices could exist remains controversial.
The future
So there are many reasons why we might never witness a technological singularity. But even without an intelligence explosion, we could end up with machines that exhibit super-human intelligence. We might just have to program much of this painfully ourselves. If this is the case, the impact of AI on our economy, and on our society, may happen less quickly than people like Hawking fear. Nevertheless, we should start planning for that impact.
Even without a technological singularity, AI is likely to have a large impact on the nature of work. Many jobs, like taxi and truck driver, are likely to disappear in the next decade or two. This will further increase the inequalities we see in society today. And even quite limited AI is likely to have a large influence on the nature of war. Robots will industrialise warfare, lowering the barriers to war and destabilising the current world order. They will be used by terrorists and rogue nations against us. If we don’t want to end up with Terminator, we had better ban robots in the battlefield soon. If we get it right, AI will help make us all healthier, wealthier and happier. If we get it wrong, AI may well be one of the worst mistakes we ever get to make.
Read more: We’ve seen the future, and it will blow your mind
New Scientist is 60 years old this week, so how will the next 60 play out? We look forward to how science and technology might upturn the world as we know it
This article appeared in print under the headline “What if… We create human-level artificial intelligence?”
