But here is another important observation. Intelligence has never been the last point of evolution, something to notice. Instead, it evolved from a myriad of solutions to a variety of challenges that allow living things to survive and take on future challenges. Intelligent current high points in an ongoing and open process. In this sense, evolution usually ends with the way people think from algorithms.
This was reflected in this open-ended, seemingly aimless sequence of challenges posed by Poet, which Clooney and others believe could be the new kind of AI. For decades, AI researchers have tried to create algorithms to mimic human intelligence, but the real progress could come from creating algorithms that try to mimic open-problem solutions to evolution – and sit back and watch what emerges.
Researchers are already using machine learning on their own, training it to find solutions to some of the most difficult problems in the field, such as how to build a machine that can learn multiple tasks at once or deal with situations they have never encountered before. Some now think that adopting this method and running with it may be the best way of artificial common sense. “We can start an algorithm that doesn’t initially have much intelligence inside of it, and see it potentially fully bootstrapped at AGI,” Clooney said.
The truth is that for the time being, AGI remains a fantasy. But this is mainly because no one knows how to make it. The advances of AI are fragmented and man-made, usually consistent with existing strategies or algorithms, which give a growing leap in performance or accuracy. Knowing what you are looking for or how many blocks you will need, Artificial has identified these efforts as an attempt to invent building blocks for artificial intelligence. And it just started. “At some point, we have to take on the Herculean task of keeping everyone together,” he says.
For us, it is a paradigm shift to ask AI to find and assemble these building blocks. It says we want to build an intelligent machine, but we don’t do anything to look like it, whatever works is just our day.
Even though AGI has never been achieved, self-learning approaches can change what makes different types of AI. Clooney says the world needs more than a very good go player. Creating a supermarket machine for him means creating a system that discovers, solves, and then invents its own challenges. A small glimpse of this in the POET activity. Clooney imagines a machine that teaches the bot to walk, then play hopscotch, then probably go. “Then perhaps it learns the math puzzle and begins to discover its own challenges,” he says. “The system is constantly evolving and the limitations of where the sky can go.”