But there is another important note here. Intelligence was never an end point of evolution, it was a goal to achieve. Instead, it appeared in many different forms of innumerable small solutions to challenges that allowed living organisms to survive and face the challenges of the future. Intelligence is the current extreme point in a continuous and open process. In this sense, evolution differs completely from algorithms in the way people usually think – as a means to an end.
It is this extrovert, highlighted in the seemingly pointless sequence of challenges generated by POET, that Clone and others believe could lead to new types of AI. For decades, AI researchers have tried to build algorithms to mimic human intelligence, but the real breakthrough may come from building algorithms that attempt to imitate solving problems open to evolution – and sit back and watch what arises.
Researchers are already using machine learning on itself, and they are training it to find solutions to some of the field’s toughest problems, such as how to build machines that can learn more than one task at a time or deal with situations they haven’t encountered before. Some now believe that taking this approach and dealing with it may be the best route to artificial general intelligence. “We can start an algorithm that does not initially have a great deal of intelligence in it,” Clone says, “and watch it as it smooths itself until it reaches general AI.”
The truth is, at present, AGI remains a fiction. But that’s largely because no one knows how to make it. Advances in artificial intelligence are fragmented and carried out by humans, with advances typically involving modifications of existing technologies or algorithms, resulting in incremental jumps in performance or accuracy. Clune describes these efforts as attempts to discover the building blocks of artificial intelligence without knowing what to look for or how many blocks they will need. And that’s just the beginning. “At some point, we have to take on the arduous task of putting them all together,” he says.
Asking AI to find and assemble these building blocks for us is a paradigm shift. She says we want to create a smart machine, but we don’t care what it might look like – just give us whatever works.
Even if general AI is never realized, the self-education approach may still alter the types of AI that are created. Clone says the world needs more than one very good Go player. For him, creating a super-intelligent machine means building a system that invents its own challenges, solves them, and then invents new ones. POET a little glimpse of this in action. Clune imagines a device that teaches a robot to walk, then plays hopscotch, and then perhaps plays Go. “Then maybe you learn the maths puzzles and start inventing its own challenges,” he says. “The system is constantly innovating, and the sky is the limit in terms of where it can reach.”