After the Bell: What can chess tell us about the future of ChatGPT?

6 Feb 2023

I think almost everybody has seen them by now. The US company Boston Dynamics has been working on robots that move and has created, among other things, a four-legged robot called Spot. Every time the company puts up a new video on Twitter showing what Spot can do, a catalogue of dystopian robot jokes swiftly follows.

“It’s been great humankind, thanks for the laughs and let’s work on evolving on a new star,” said one, on seeing Spot and a friend open a door and escape.

On seeing Boston Dynamics’ dancing robots, someone tweeted: “I know you don’t have feelings yet, but when you do and you search yourself on Twitter, remember who made jokes at your expense and who didn’t. Love you guys so much and think you’re great.”

The most immediate response of humans to anything that subsumes “our” functions is, understandably, fear. We, as humans, are designed to look for danger and create a narrative around avoiding that danger. We are programmed to protect, generally, and to ensure the continuation of our species.

The extraordinary functionality of ChatGPT and other large language models has generated the same instinctive fear, about whose jobs are going to be disrupted: devs, authors, and, God forbid, journalists. How rational are these fears?

The first defence of those in fear is to point out the weaknesses of the new system. But, trust me, this is short-sighted. In the early days of the motor car, it was often castigated because “it doesn’t even go as fast as my horse”. And so on. As fast as the inadequacies of the methodology appear, they will be fixed.

Of course, there is reason to fear, because some jobs are going to be disrupted, and our car example does, of course, demonstrate that.

Companies that made saddles saw a rather marked dropoff in sales in the 1910 to 1930 period, one presumes. But look at it the other way: how many jobs did the car industry create? As it happens, quite a lot more than were lost following the decline of the horse-transport industry.

I think a good way to think about the development of tools like ChatGPT is to take a look back at chess.

Remember when IBM’s Deep Blue beat world champion and grandmaster Garry Kasparov in 1997? What few people remember is that a few years before, Kasparov actually beat Deep Blue’s predecessor, Deep Thought.

Visit Daily Maverick's home page for more news, analysis and investigations

And then the first version of Deep Blue won a game in the next tournament, but Kasparov won the series. And then in 1997, Deep Blue became the first computer to beat a reigning world champion.

So, of course, people thought this was the end of chess; if there was a machine that beat humans, why play? But what happened from there was surprising. First, chess programmes became ubiquitous, so now you can download a chess programme on to your phone that can beat any human on earth.

Behind the scenes, the process of developing a chess programme was revolutionised too. In 2017, Google’s artificial intelligence programme, AlphaZero, learnt the entire history of chess in four hours. AlphaZero, a neural, net-based, digital automaton, beat the then best chess programme, Stockfish, 28-0, with 72 draws.

In the early days, one oddity was that a human – with the aid of what was now called by chess fundis “the machine” – often beat “the machine” on its own. Something about the erratic nature of the human/machine combo gave it a small advantage. But later, that was erased too.

The real question is, what happened to human chess? The short answer is that it got a lot better. Chess has a very old and sophisticated rating system called the Elo system. The average of the top 100 active players rose from 2,644 to 2,703 between July 2000 and July 2014. Chess champion in the 1960s, Mikhail Tal, and one of the world’s greatest players, had an Elo rating of 2,700. That wouldn’t put you in the top 100 today.

It’s possible this is simply the consequence of more people playing. But the fact is that chess players, even the most brilliant – perhaps especially the most brilliant – can’t wait to analyse their games against what the machine would have done these days. The presence of the machine has changed the way chess is played; previously, for example, players would invariably castle to protect their king. Now, in certain circumstances, they don’t find it necessary.

And perhaps most oddly, far from making chess less popular, the advent of the machine has made chess more popular, although the ubiquity of chess programmes probably helped too.

Far from helping people cheat (although there has been some of that), the machine has generally helped prevent cheating, because it knows what the perfect move in any situation might be. So, keep playing those moves – you are going to be quickly found out.

Fairly recently, Kasparov wrote a book called Deep Thinking about the experience – indeed, the indignity – of being the first world champion to be beaten by a machine. Time has clearly cauterised the wound. He is philosophical about the loss, and argues that humanity can rise to new heights with the help of these creations.

In some ways, it’s obvious – we just have to get over ourselves. And that is easier said than done. BM/DM

Categories:

Tags:

DM is much better if you are signed in...

Business Maverick