GTO Poker Theories - Moravec’s Paradox

What we thought would be easy/tough for AI turned out to be the opposite, and what that means for poker.

One of the real gifts poker has given me is that it has been a great jumping-off point to learn things from other disciplines like economics, AI, psychology, and Game Theory. So here is a series of articles where I bring some of the most interesting things I have learned from other subjects outside of poker which are applicable to this game we know and love.

In the early days of artificial intelligence, many researchers assumed that advanced reasoning would be the hardest task for computers. They expected activities like complex problem solving, playing chess at a high level, or doing difficult maths to require enormous processing power. At the same time, they thought skills such as walking, recognising objects, or understanding spoken language and emotions would be easier to program. It turned out the opposite was true.

This surprising finding is known as Moravec’s Paradox. Things humans do without thinking, like balancing while we walk or picking out a friend’s face in a crowd, are extremely hard for computers to learn. By contrast, tasks many people find difficult, like working through long calculations, can be done quickly and easily by a computer. The reason is that our everyday physical and sensory skills are the result of millions of years of evolution, while logic and abstract thought are much newer in human history. Computers are designed for the latter type of task, so they naturally excel at it.

Moravec’s Paradox shows that intelligence is not just about abstract reasoning. It also includes sensing the world, moving through it, and responding in real time. These “simple” skills for humans are still some of the biggest challenges for AI and robotics. Even with modern advances, teaching machines to match human abilities in these areas is slow and complex, and the gap is likely to remain for a long time.

The soft skills

From a poker perspective, we have already seen that the technical side of the game has been mostly solved, or at least it is on its way there. 

Thanks to solver technology, we have a sense for what the optimal line in a hand is. With every year, the speed at which we can 'solve' a hand is getting much faster (worringly so). 

It seems, however, that live poker is far from solved if Moravec's Paradox is anything to go by. It seems like soft poker skills, like table talk, spotting tells or sensing if an opponent is on tilt, are something a computer may never be able to do. We may never, for example, be able to wear a pair of smart glasses and pick up when our opponent is weak, in the way that a seasoned live professional may be able to. 

This may explain the global trend of live poker numbers accelerating since the pandemic. The live game is not immune to the threat of solver technology, but there are aspects of the live experience that only humans can excel at, which an app could not help with.  

What theories from outside of poker have helped your game? Let us know in the comments.