Negotiators needn't fear being surpassed by AI - as Garry Kasparov can explain

This article was first published on Forbes.com and is reprinted with permission.

Predictions of AI doom have been wrong before.

When reigning World Chess Champion Garry Kasparov accepted a challenge in the 90s to play against IBM’s Deep Blue supercomputer, many in the chess community were aghast. They feared a victory for the machine would deliver a fatal blow to their 1,500 year old game.

At the time, computer technology was spreading globally and many chess veterans reasoned that if the best human player was shown to be inferior to a machine then interest in playing would wane and sponsorship would disappear. They urged Kasparov not to take part in the spectacle.

Against that advice, Kasparov was eventually defeated by the computer in 1997. He became the first person to lose their job to AI, as he now jokes.

Yet, today, chess is far more popular than ever before. It’s thriving in the age of machines with top players emerging from countries that didn’t even have wide interest in chess back then. The current World Chess Champion is Ding Liren from China where the game has grown exponentially in popularity over recent decades. And sponsors have never been more keen to invest in human players. Magnus Carlsen, until recently World Chess Champion, has amassed a fortune from playing that’s estimated in the tens of millions of dollars.

Understanding how chess flourished, not just in spite of machines but because of them, can help us better understand how AI is transforming our wider world today as it rapidly learns to outperform human capabilities in other areas, far more consequential than a game of chess.

Like all technology, AI is not inherently good nor bad. Nor is it neutral. It will have a profound impact in a complex myriad of ways. Many concerns today about the dangers of AI are entirely valid but that’s why it is even more vital that we understand and embrace the positive potential of AI in order to steer its wider development further in that direction.

As it happens, the concerns of weary chess veterans back in the 90s about the wider impact of the machine age were not entirely misplaced, at least in terms of their own narrow interests.

In the years that followed Deep Blue’s victory, there was a rapid upheaval of chess rankings with long time leaders suddenly usurped.

Before the spread of computers, you’d need much more than talent to succeed in chess. You’d need to be among the lucky few whose talent would be recognized and nurtured by existing experts in the game with whom you could hone your skills.

Yet, suddenly, computers could enable anyone to play endlessly against experts, both artificial and online around the world. Not only that, budding new players could use the internet to follow games taking place that they’d never be able to attend. They could then also get immediate and detailed analysis.

And a lot of new tactics needed analyzing. Because computers didn’t just learn to play well. They played differently and shaped the human game in the process. Despite centuries of human play, computers could still innovate new tactics and approaches by crunching the vast amounts of data about gameplay they were consuming. This inspired human players to rethink and adapt how they play with each other.

Machines didn’t replace humans in chess. In the professional game, humans who embraced machines replaced humans who didn’t. But the overall impact was an enormous net positive.

The same pattern repeated itself more recently in the even older and more complex game of Go when Google Deep Mind’s AlphaGo swept aside some of the game’s top players.

Starting in 2016, the first Go player to suffer defeat to a machine was European Go champion Fan Hui. Yet his ranking vastly improved afterwards, which he attributed to AlphaGo helping him see the game differently. Next to be stunned was the popular Korean player Lee Sedol who said afterwards that the computer’s style was so different that it made him realize he needs to study Go more. Finally, the world’s top ranked Go player Ke Jie suffered defeat in 2017. It was all too much for dismayed Chinese state censors, which prohibited further coverage of AlphaGo.

As DeepMind continued to develop AlphaGo, it first taught itself similar opening moves to top human Go players but eventually discarded them in favor of its own entirely novel opening strategies.

At the same time, human go players were given access to the AI data it was crunching and they used it to develop new strategies to play against other humans. The change was quantifiable. As reported by Scientific American, an analysis of matches between human players showed relatively stagnant gameplay over previous decades but which was suddenly transformed in the years following AlphaGo’s emergence, both by novel moves played earlier in games and higher decision quality.

It would be understandable if Garry Kasparov himself bore a grudge against AI.

Instead, over the decades since, Kasparov has grown more enthusiastic about how machines have enhanced and democratized the human game, as he writes in his book, Deep Thinking.

Kasparov also now sees parallels with the positive potential that AI can have on our wider world if developed in the right way and led by free countries that can shape the rules in the interests of everyone.

Beyond the media spectacle, classic games have served as fascinating tests for the development of AI ever since primitive machines were first programmed to beat humans at noughts and crosses. The real world is more complex, however, and AI has no inherent desire to play against us.

Take negotiation, for example, as one of the most fundamental human skills required to induce cooperation and create value on which the global economy depends. Sadly, humans tend to be quite bad at it. Not only do we have all kinds of cognitive biases that cloud our better judgment, but there is far too much data to analyze for every potential deal around us.

Just as chess and Go moves quickly add up exponentially in complexity, the terms to a fairly typical contract negotiation with a limited number of terms can have quintillions of possible outcomes. So it’s no surprise that AI has now surpassed human capabilities here too with the emerging field of autonomous negotiations.

Once again though, the way AI plays this ‘game’ is quite different from humans. Humans have a tendency to see negotiation as a zero sum game with a winner and a loser. But AI can be tasked to win by extracting the most value from a deal so it does that by making sure both sides can create more value together. In mathematics, that’s known as Pareto efficiency.

In other words, it bakes a bigger pie in order to get a bigger slice of it.

This is beyond even our concept of a win-win outcome. Take a classic tale often told in negotiation training. Two kids are arguing over who gets the last orange. The parent tells them to split it in half. That could be categorized as a win-win. But what if one child wanted to eat the orange slices and the other needed the orange peel for a cake recipe? If that information was known when negotiating the orange, both sides could have gained everything they wanted.

In the real world, these kind of missed opportunities for better deals are all around us in more complex ways so if AI can harness more information to conduct autonomous negotiations at scale then this is just one more way that AI can significantly raise the world’s GDP and provide a net positive to humanity, while teaching us new skills in the process.

According to a recent IMF study, 40% of jobs globally will be impacted by AI. That rises to 60% in advanced economies. Behind the headlines though, those figures break down to roughly half of impacted jobs at risk of being replaced by AI while the other half will likely be transformed by AI acting in a complimentary way. And that’s before we consider the new jobs that would be created by AI, as has happened with previous technologies when humans unlock value never before possible.

When we walk into an elevator, we no longer expect a human operator. When we board a plane, we expect a pilot but do hope their autopilot is working too. And any busy elevator or plane these days will be filled with people doing jobs that didn't exist when those things were invented.

What’s happening with AI is really a continuation of longer term trends at a far more rapid pace, which is why it’s helpful to look back into the past to understand where it’s heading.

As in chess previously, it’s not necessarily AI that will take your job. It’s humans using AI that will be taking jobs from humans not using AI. But there is the potential for an enormous net benefit for everyone if we recognise the emerging opportunities and begin adapting for them.