If games teach us to be human, what are computers learning from them?

Time will tell whether it's a smart move for scientists to use games like chess and Go to teach AI machines human intelligence

27 May 2018 - 00:00 By aspasia karras
subscribe Just R20 for the first month. Support independent journalism by subscribing to our digital news package.
Subscribe now
Walter Wisby, aged eight, playing a game of chess with T Whiltard, aged 91, in Cheltenham in August 1913.
Walter Wisby, aged eight, playing a game of chess with T Whiltard, aged 91, in Cheltenham in August 1913.
Image: Hulton Archives/ Getty Images

My childhood game playing can be divided into two categories: games at school and games at home. School games were largely organised around traditional pursuits. Marbles in marble season, elastic during elastic season, Running Red Rover in Running Red Rover season and so on.

The seasons would start and end by general consensus, which would happen by strange childhood osmosis. Without a word spoken, everyone would know that it was the time of year to start playing marbles. The wind would change and without question your marble bag would be fished out from under the bed or the back of the cupboard.

Your goonies, your milkies, your tiger's eyes, spit polished, closely fondled and intensively investigated for chips. The counting was crucial. Lots of counting and counter counting of marbles took place. And then, in a cloud of dust, the season would begin.

Small tacticians and gamblers would gather in the playground in neat trading rows and set up their wares to take their chances and hopefully all of their friends' marbles

Small tacticians and gamblers would gather in the playground in neat trading rows and set up their wares to take their chances and hopefully all of their friends' marbles. There were constant side deals going on where you might swap several bog standard marbles for some delicious marble rarity. The sheer delight of victoriously claiming someone's little pile of marbles with bare-knuckled marble-rolling skill still resonates.

And then, just as you had mastered the fine art of barter and had started to accumulate much marble wealth, the season would shut down and you would instantly abandon the marble bag at the bottom of your school bag - a forgotten ballast weighing down your dreams of marble domination till next year.

Everyone moved on ruthlessly and single-mindedly. The old games forgotten, like last year's dead silkworm moths. Now the new season worms would hatch and with them feats of extreme jumping with elastic songs. The following week would bring the constant threat of concussion as you ran full steam into the human chain that ranged across the field, chanting "Red Rover, Red Rover, something something come over".

The month after that it was the avaricious collecting of stickers and a constant state of penury as you threw what few coins your parents gave you as pocket money after your mad obsession - fingering the sheets of stickers at the stationery shop with deep longing. The ones with the mobile eyes were the best. The very best. The bestest.

LAST SEASON'S RULES

Kissing Catches, Spotlight and Spin the Bottle all also started spontaneously one year - these games were entirely unprecedented.

Variations of Kissing Catches sprang up, including something called Pass or Fail, which introduced terrible judgment into the previously unselfconscious world of games. Best to retreat home, where the kid brother was still operating under last season's rules. Which brings me to home games.

Fishing clickets out of my brother's nose, where he would sometimes lose them, was playing doctor-doctor on a very sophisticated level

Fishing clickets out of my brother's nose, where he would sometimes lose them, was playing doctor-doctor on a very sophisticated level. Magnets are amazing things even with all that skin and cartilage in between. Interestingly, he became the doctor in the family - clearly inspired by those early childhood operations.

My aspirations were very differently expressed in games of Career Girl, Career Girl, which involved much flying about to New York and Paris in a helicopter piloted by a certain Peter of our acquaintance. He was not averse to being called TC after the dashing helicopter pilot in Magnum PI.

I spent hours constructing perfect magazines - glorious scrapbooks filled with everything a person of substance might want to read. I would pen all the articles and purloin the pictures from other publications - with a strong preference for portraiture. Primarily portraits of John Stamos. Who would never have failed me in Kissing Catches. Just saying.

TORTURE, TORTURE

I have heard arguments that strongly suggest that the games of childhood are merely training for life. Children play in order to learn how to be human. All the bartering, collecting, negotiating, fantasising and concussing are just ways of patterning the future, and modelling ways of being in adult society.

Which is why I am not dwelling on Torture, Torture, which was a firm childhood favourite, played with the same girlfriends of Career Girl, Career Girl. For obvious reasons they shall remain nameless. The worst torture was enforced bathing, after squishing of toes in mulberries. Terrible stuff.

In the interest of my future self and to probably distract me from Torture, Torture, my father valiantly tried to introduce chess. But my strategic foresight was limited to a maximum of two moves ahead of the current state of play.

So no budding Kasparov then. I probably got more out of the pure mindless simplicity of tennis bats and warships on our antique Atari. That was then - the Atari has played a lot of games since those halcyon days of computing innocence.

Go, an ancient Chinese game.
Go, an ancient Chinese game.
Image: Hulton Archives/Getty Images

Kasparov must have felt like a babe in the woods when Deep Blue, the IBM super computer, beat him hands down in 1997. And the IBM Deep Blue shape-shifted into AI super-genius Google Alpha Go, because everything has to grow up sometime.

This latter-day super AI brain pitted itself against Go, the ancient Chinese game, because it was a much bigger algorithmic and strategic challenge.

Go has far more move options and is apparently a game of street-smart, deep-level human intuition intended as a strategic life-enhancing game, teaching players crucial lessons to be applied when adulting.

Which makes me wonder what Alpha Go has learnt, now that it has consistently beaten the teenage global Go champion Ke Jie. Alpha Go had played itself millions of times before it pulled its spectacular new moves against Jie, who took it to heart, and equated the machine with God.

Jie was a little disheartened and said he would never play Alpha Go again. Apparently he has changed his mind lately, because which teenager wouldn't want to pit himself against the ultimate adult - God?

Is the fact that scientists are using games to teach AI machines human intelligence the obvious route to teach them how to be human? And now that they are outplaying us at our own games intended to teach us how to be humans, what comes next? I just wonder if Alpha Go knows when spontaneous game-changing season is up?


subscribe Just R20 for the first month. Support independent journalism by subscribing to our digital news package.
Subscribe now