AI learns to create itself
But there is another crucial observation here. Intelligence has never been an end point for evolution, something to aim for. Instead, it emerged in many different forms, countless tiny solutions to challenges that enabled living things to survive and meet future challenges. Intelligence is the current culmination of a continuous and open process. In this sense, evolution is quite different from algorithms from the way people generally view them – as means to an end.
It’s this openness, seen in the seemingly aimless streak of POET-generated challenges, that Clune and others believe could lead to new types of AI. For decades, AI researchers have tried to create algorithms to mimic human intelligence, but the real breakthrough may come from building algorithms that attempt to mimic evolution’s open problem solving – and sit down to watch what emerges.
Researchers are already using machine learning on itself and training it to find solutions to some of the most difficult problems in the field, such as making machines capable of learning more than one task at a time or face situations they have never encountered before. Some now believe that taking this approach and following it might be the best path to general artificial intelligence. “We could start an algorithm that initially doesn’t have a lot of intelligence, and watch it build up to potentially AGI,” says Clune.
The truth is that for now, AGI remains a fantasy. But that’s largely because no one knows how to get there. Advances in AI are piecemeal and made by humans, with advancements typically involving adjustments to existing techniques or algorithms, resulting in incremental leaps in performance or accuracy. Clune characterizes these efforts as attempts to uncover the building blocks of artificial intelligence without knowing what you’re looking for or how many blocks you’ll need. And that’s just the beginning. “At some point we have to take on the Herculean task of bringing them together,” he says.
Asking AI to find and put these building blocks together for us is a paradigm shift. This means that we want to make a smart machine, but we don’t care what it might look like – just give us what works.
Even if AGI is never achieved, the self-learning approach can still change the types of AI created. The world needs more than a very good Go player, says Clune. For him, creating a supersmart machine means building a system that invents its own challenges, solves them, and then invents new ones. POET is a little glimpse of this in action. Clune imagines a machine that teaches a bot to walk, then play hopscotch, then maybe play Go. “So maybe he learns math puzzles and starts inventing his own challenges,” says -he. “The system is constantly innovating, and the sky is the limit in terms of destination.”