Throughout history – or at least, recent history – there have always been those signpost moments where it seems that technology has outstripped what humans can do in a certain field. Back in 1997, for example, there was the famous battle between chess grandmaster Garry Kasparov and the IBM computer Deep Blue. Kasparov was considered among the best chess players in history, but Deep Blue succeeded – with relative ease.
The Deep Blue chess game, which has been immortalised in a fascinating documentary, Game Over: Kasparov and the Machine. It’s really worth watching as it not only chronicles a milestone in the evolution of computers, but it also poignantly shows Kasparov dealing with his own limitations. There have been other milestones like this one, including another IBM supercomputer – WATSON – thrashing the Jeopardy champions on the US quiz show in 2011. IBM doesn’t have the monopoly on these supercomputers, of course: Google got in on the action when DeepMind AlphaGo defeated the world’s best Go player in 2017.
AIs have mastered games involving structured data
In the 2020s, of course, it does not seem much of a surprise that a supercomputer can learn millions of pieces of general knowledge or learn how to play chess. Even if chess is a complicated game, it is understandable how a computer can master it. The moves, by Kasparov, are the kind of logical human thoughts that a computer can understand. It might take a lot of time to set Deep Blue on its way, but you can appreciate how IBM might succeed. Go, while more complex than chess, also has logical move sets, so it feels natural that Google’s AI beat the best human player.
But what about the world of illogical? Can computers succeed in that world too? Something like sports predictions and sports betting, which are both inherently illogical (we are using this term broadly to signify randomness and creativity) in their outcomes, are now popular subjects for tech companies who want their AIs to conquer more complex human activities than board games and quizzes.
Computers still struggle with unstructured data
Can a computer beat an expert sports betting tipster? The short answer is not yet. But it’s interesting to muse upon the reasons why.
If we look at the example of the NFL. There’s a massive industry around betting and fantasy sports that involve making predictions for NFL games. If you consider something like 888 Sport’s NFL picks straight up, the experts analyse weekly games in the NFL (the new season is just about to get underway). But these experts, like all good sports tipsters, will look at two main areas.
The first is essentially data and precedence – how teams and players have performed in the past, how they have performed under similar circumstances, and so on. The second area they will look at is a bit more difficult to define. It’s to do with emotions and intuitions, knowing, for example, that a player’s head won’t be in the game due to off-field issues – that type of thing. We call the former structured data, and the latter is called unstructured data.
As you might have guessed, computers can analyse structured data much better than humans can – it’s not even close. Whatever was being fed into Deep Blue and WATSON is structured data. But humans can understand unstructured data better than computers – at least we can right now. There is a consensus that the gap is narrowing, and pretty soon, the betting experts – like Garry Kasparov – will be left behind.
Will that mean a future where we can all get rich by getting betting tips from an AI like WATSON? Not necessarily. Sport will remain the same, and that means there is always unpredictability intertwined with the game. Even the most sophisticated AIs will not be able to predict that, for example, Simone Biles will remove herself from the Olympics team or that Roger Federer will require surgery. But when it comes to the analysis of the information available, computers will be better placed to make informed predictions than the rest of us.
Nerd Comments