Fourteen years into the 21st century, there have been no little green men, no meal-replacement pills, no flying automobiles, no space odysseys. But as big-budget plans to model the human brain prove, artificial intelligence remains in the realm of the barely plausible. In its most literal sense, AI exists already: encoded and executed, endowed with sensors, lenses and microphones, connected to the internet, and stuck in your pocket. But how intelligent does a machine have to be before our worst nightmares come true? Intelligent enough to pass a Turing test? Intelligent enough to nuke the human race? And/or intelligent enough to be self-aware, and thus real by Cartesian standards? Apocalypse notwithstanding, that’s the threshold we’re really interested in: artificial consciousness, artificial free will, and artificial bodies for that matter.
But spending all the world’s neuroscience dollars on a supercomputer simulation of the brain’s neuronal connections will reveal less about AI and more about human stupidity. PZ Myers writes, “We aren’t even close to building such a thing for a fruit fly brain, and you want to do that for an even more massive and poorly mapped structure? Madness!” If the IT resources exist to simulate 90 billion neurons and 100 trillion connections between them, mediated by dozens of different neurotransmitters and organized into highly specialized networks, there’s still no reason to expect intelligence to emerge or a ghost to glom on to the machine. The scientific consensus is that there’s still much to learn about the brain, and this will only be achieved through less grandiose and far-fetched research.