Artificial Intelligence is one of those terms that’s been tossed around for a long time, percolating into pop culture for decades now, as cautionary tales like the movies Terminator or War Games.
Now it’s about to get real. Ray Kurzweil, the futurists inventor and Google bigwig, has identified what he calls history’s Law of Accelerating Returns. It’s not complicated: more advanced societies have the ability to innovate faster than less advanced societies.
Think of the progress that’s been made in the last two centuries – going from horses to cars to spaceships, from pen and paper to the telegraph to the Internet. The pace of change has been extraordinary compared to all the previous years of human history.
And it’s continuing. We are on the cusp of innovations now that will further transform our economy and society.
Artificial intelligence will be the most important of these changes.
Tim Urban, in a series of very good posts, explores the rise and consequences of AI. He breaks down AI into three distinct buckets:
1) Artificial Narrow Intelligence (ANI): Sometimes referred to as Weak AI, Artificial Narrow Intelligence is AI that specializes in one thing, like being a chess champion.
2) Artificial General Intelligence (AGI): Sometimes referred to as Strong AI, or Human-Level AI, Artificial General Intelligence refers to a computer that is as smart as a human across the board—a machine that can perform any intellectual task that a human being can.
3) Artificial Superintelligence (ASI): Artificial Superintelligence ranges from a computer that’s just a little smarter than a human to one that’s trillions of times smarter—across the board.
This simple graphic tells the very powerful story of what AI will do relative to the rate of human progress:
We are on the edge of a remarkable inflection point. This is one time when the hype cycle of new technology doesn’t really do justice to the potential for change. It won’t be incremental. It will be revolutionary.
Some call reaching the superintelligent phase the “singularity” – the point in time when computers become both aware and exponentially smarter than their human creators. Kurzweill says the human brain will be reverse-engineered by 2029. By 2045 he says we will achieve the singularity. “The whole point of it is to extend our reach.”
Not surprisingly, a lot of people are worried about the rise of machines that are smarter than humans. Elon Musk, Stephen Hawking and Bill Gates have all expressed concern about AI, with Musk comparing it to “summoning the demon” we can’t control.
There is even an NGO called the Campaign To Stop Killer Robots, an international coalition that wants to preemptively ban fully autonomous weapons. Armies already have drones that can be guided by a soldier on a videogame-like interface from thousands of kilometres away. The next step is sentient drones that don’t need the humans.
Of course, not everyone fears AI. Legendary Internet pioneer and venture capitalist Marc Andreesssen is bullish on the technology, arguing it will be positive for the future of humanity. He says that those fearful of it are falling into the myth of Prometheus. (He talks about it in this recent interview with podcaster Tim Ferriss).
Want to build some AI yourself? It’s easier than ever. Google has even made its TensorFlow available as an open source software library for machine intelligence.
What does AI mean for our workforce? Why hire a human when a guided robot works better, faster and cheaper? Technological innovation is a net creator of jobs, but the impact of AI may be different.
At the consumer level, rudimentary AI is already all around us today, in the form of Apple’s Siri, Microsoft’s Cortona or Amazon’s Echo or in many of the features found on the web and in your smartphone.
But those are baby steps, we haven’t reached the second and third stages of AI yet – but smart people are working hard on creating technology that will be smarter than them. A big part of this is computational power – giving a computer the power of the human brain.
How fast is that? Ray Kurzweil frames this in total “calculations per second” (CPS) for the human brain and arrives at the figure 10 quadrillion cps. That’s a lot.
But Tim Urban points out that China has already constructed a supercomputer that beats that number, able to reach 34 quadrillion cps. Of course it cost $390 million to build, takes up 720 square meters of space and uses 24 megawatts of electricity. So it will be a while before it gets shrunk down to fit on your wrist or in your pocket.
Still, just as we’ve moved from computers the size of a building to those that fit in a smartphone, computational power won’t be the barrier for long. The bigger challenge might be making AI smart like a human, with emotional intelligence and able to discern and learn. And here’s where it gets really trippy.
One way to do this would be “whole brain emulation” – slicing a real brain into thin layers, scanning each one and then building out a 3-D model. That could create a computer capable of doing anything a human brain is capable of doing. Even trippier, the process could eventually capture the consciousness of a real person and place that into a machine. So while our bodies might die, in this scenario humans could “live” forever in mechanical hosts.
Soak on that for a moment.
So AI could mean people are crushed under the boot of superintelligent robot overlords who no longer need our services, or we might live in a beautiful world where the problems of humanity are solved and no one ever really “dies.”
It should be fun to watch.