Artificial intelligence is learning how to create itself

However, here is another crucial discovery. Intelligence is never the end of evolution, but the goal. Instead, it comes in many different forms, from countless tiny solutions to challenges that allow living things to survive and meet future challenges. Intelligence is the highest point in the ongoing open process. In this sense, evolution is completely different from what people usually think of as algorithms, which are a means to an end.

It is this openness that convinced Clune and others that a series of goalless challenges that may lead to POET has led them to produce new AI. For decades, artificial intelligence researchers have been trying to build algorithms that mimic human intelligence, but the real breakthrough may come from building algorithms that try to imitate evolutionary open-ended problem solutions, and then sit down and observe the results.

Researchers have used machine learning on themselves and trained them to find solutions to the most difficult problems in the field, such as how to make a machine learn multiple tasks at once, or how to deal with situations that have never been encountered before . Some people now think that adopting this method and putting it into practice may be the best way to artificial intelligence. “We can start an algorithm that doesn’t have much intelligence inside at first, and then observe that it may continue to bootstrap until it may reach AGI,” Clune said.

The fact is, for now, AGI is still a fantasy. But this is mainly because no one knows how to do this. Advances in artificial intelligence are sporadic and are completed by humans. Advances usually involve adjustments to existing technologies or algorithms, leading to a leap in performance or accuracy. Clune describes these efforts as attempts to discover the basics of artificial intelligence without knowing what you are looking for or how many blocks you need. This is just the beginning. He said: “At some point, we have to take on the arduous task of putting them together.”

Asking AI to find and assemble these building blocks for us is a paradigm shift. This means that we want to create an intelligent machine, but we don’t care what it looks like-just give us any feasible method.

Even if AGI has never been implemented, the self-study method may still change which type of AI is created. Klein said that the world not only needs an excellent Go player, but also more. For him, building a super machine means building a system, inventing one’s own problems, solving them, and then inventing new ones. POET is a glimpse of the actual operation. Clune imagines a machine that teaches robots to walk, play hopscotch, and then play Go. He said: “Then it might learn math puzzles and start to invent its own challenges.” “The system continues to innovate. As far as its possible development is concerned, the sky is the limit.”