Elon Musk: Killer Robots Will Arrive Within Five Years
Elon Musk is at it again. In the last year, he's made a series of incendiary claims about artificial intelligence: that it is a greater threat to humanity than nukes, that it threatens to turn humanity into "biological boot loaders," and that it is akin to summoning a demon. Now, he's claiming that the advent of killer robots is much more imminent than we would have imagined: "The risk of something seriously dangerous happening is in the five year timeframe."
The post, hosted on Edge.org, was deleted shortly thereafter, but here's a screenshot of the full comment from Reddit user Buck-Nasty:
In this post, Musk asserts that robots will become hostile towards humanity in "ten years at most." He claims that AI companies are worried as well, at least behind closed doors, and that a "digital superintelligence" is on the horizon. Superintelligence is a version of the technological singularity in which an AI becomes more intelligent than humans, often because it is advanced and self-aware enough to make improvements to itself.
Edge purports itself to be the living document of the The Third culture, which "consists of those scientists and other thinkers in the empirical world who, through their work and expository writing, are taking the place of the traditional intellectual in rendering visible the deeper meanings of our lives, redefining who and what we are." As Buck-Nasty pointed out, they only allow commenters who have been invited to create an account, so it would have been difficult for anyone to impersonate Musk.
Just how justified are Musk's concerns? It's impossible to tell for certain, but there are currently conflicting notions of our progress in the artificial intelligence among experts. Some experts, such as Oxford University's Nick Bostrom, insist that superintelligence is a very real threat to humanity, possibly an even bigger threat than global warming. But there are just as many experts who insist that our current AIs have trouble with even the most basic tasks, with one study asserting that one of the most advanced AIs at the moment has the IQ of a four-year-old.