Scientists Think They Know the Exact Year Computers Will Render the Human Brain Obsolete


This week, the greatest minds in science and technology pleaded with the world to prevent an artificial intelligence "arms race" — an apocalyptic scenario in which terrorists would have access to highly advanced weapons like killer robots.

The issue arose back in April when representatives from Human Rights Watch and Harvard Law School wrote a paper calling for the United Nations to ban "killer robot" production until regulation and legal stipulations — for instance, who's at fault when a robot shoots an unassigned target — could be put in place.

We should be listening when some of the most trusted minds in tech, including Elon Musk, Steve Wozniak and Stephen Hawking, are warning the world about letting AI weapons go off the rails. But history and science demonstrate that their predictions for hyperintelligent computers aren't as far-fetched as they sound. Computers could soon "think" faster than human beings too.

"The problem is, when the machine realizes it can do anything and grow in terms of speed, capacity and memory, it might learn to deceive us very quickly." — Zoltan Istvan

Computers of the future: A principle called Moore's law predicts that computing will increase in power while simultaneously grow exponentially cheaper and smaller. A perfect example: In 1956, IBM needed a forklift to put a 5-megabyte hard drive, then the size of two refrigerators, on an airplane. Today, for $20, you can buy a 16-gigabyte flash drive that fits on your keychain. That's 3,200 times more data capacity than the aforementioned behemoth.

Ray Kurzweil: 'The Singularity is Near: When Humans Transcend Biology'

Ray Kurzweil's 2005 book The Singularity Is Near: When Humans Transcend Biology (available as a Google PDF) charts our modern era's computational power and compares it to a mouse brain. But in order to reach human-brain levels of processing, the way we approach the technology may need to be approached differently. Kurzweil anticipates optical computing, which operates extremely fast but needs much less power.

Thanks to a Duke University research team, which developed an "ultrafast spontaneous emission source" to set a new speed record, that technology could be closer than expected.

The $1,000 brain: Engineer Peter Diamandis used the now 50-year-old Moore's law to make eight predictions about where technology will go in the next decade. His very first prediction is that we'll have a $1,000 "human brain" — that is, a computer that can perform 10,000 trillion calculations every second, or the same as a human.

Think of any film in which an AI robot speaks. Its response time is slow because it's calculating what to say. But if the machine could process at human speeds and faster, it could respond with lighting-fast computation virtually as quick as human instinct. Which is awesome for passive, friendly applications. But what happens when robots can think faster than humans? Should we be worried?

"The problem is, when the machine realizes it can do anything and grow in terms of speed, capacity and memory, it might learn to deceive us very quickly," Zoltan Istvan, a transhumanism activist and columnist, told Mic. "We might think we have the perfect child, but then the child is in the backyard setting things on fire."

Computing speed and artificial intelligence are not the same thing. But if artificially intelligent robots could come factory-standard with processing speeds faster than a human brain, then it could theoretically make decisions and act on them before a human has any idea what's happening.

And that's a little scary.

Some sources say Moore's law isn't hurtling through technological milestones the way it used to. Some argue that it's fallen flat, while others say it's continuing to defy expectations. Optical technology, however, would grant the capacity for unprecedented computing speeds and exponential growth in a completely new area of innovation.

Meanwhile, the war against murder machines wages on, argues Bonnie Docherty, author of the paper urging the UN to halt killer-robot progress until it's properly regulated.

"The next major step at the UN will be the annual meeting in November of countries that are party to the Convention on [Certain] Conventional Weapons," she told Mic. "That's the disarmament forum we talked about in May. At that meeting, countries will decide what form future engagement with the issue will take. We are optimistic discussions will continue, and we are urging countries to agree to take the next step and begin negotiations of a new treaty to ban killer robots."