Artificial intelligence (AI) can mean a lot of things. It can include anything from digital assistants to warehouse robots. The algorithms used to power those devices are in almost everything — if it has wires and computer chips, it likely uses some form of AI.
But Scott Phoenix, the cofounder of AI company Vicarious, said we shouldn't be referring to all these things as AI.
"AI has become a very diffusely defined term that can be applied to anything," Phoenix told Tech Insider.
"They talk about it in terms of this spam filter ... or they'll talk about Google self-driving cars having AI because they can drive a car. It's very rapidly becoming a word that just means the system can do stuff you want it to."
According to Phoenix, there's only one thing that should be considered AI: an artificial being that can do all the different things a human can do, as well as a human can.
"AI, to me, really means something specific which is, given the same kinds of sensory motor inputs that a human has, from birth to adulthood, your system should form the same concepts and have the same capabilities," he said.
In fact, that's exactly the kind of AI that Vicarious, which is backed by Elon Musk, Mark Zuckerberg, and a bunch of other tech stars, is trying to build. The company, which was founded in 2010 by Phoenix and neuroscientist Dileep George, is doing something revolutionary in computer science: They want to build the world's first human-level AI.
"Vicarious is building a single, unified system that will eventually be generally intelligent like a human," Phoenix wrote in a World Economic Forum Q and A.
There's nothing available now that would be considered AI according to Phoenix's definition. Nothing even comes close. The kind of AI systems available now are very good at narrow tasks, like playing chess or buying stocks — they're a long way off from human-level AI, which would have be good an almost limitless range of things.
Philosopher Nick Bostrom surveyed 550 AI researchers to gauge when they think human-level AI would be possible. The median answer from the researchers was that there is a 50% chance that it will be possible between 2040 and 2050, and a 90% chance that it will be built by 2075.
Having taken the survey, Phoenix agrees with that timeline, though he couldn't recall what his answer was. Asked when he thinks Vicarious' own human-level AI system will be ready, he responded that most predictions about when different kinds of technology are available usually completely miss the mark.
"The goal of Vicarious is to solve this problem and work on it for as long as it takes," Phoenix said.
