On Computers and Chess and Go

On Computers and Chess and Go

The latest edition of New Scientist [October 21, 2017] reports that Google's super computer DeepMind no longer needs human modeling to devise ultra-sophisticated strategies in the ancient Japanese board game Go. This comes on the heels of myriad laments regarding the apparent impotence of human chess grand masters when confronted with computers like the newly crowned Komodo which according to ExtremeTech [March 15, 2016] "can reach an Elo rating as high as 3304 — about 450 points higher than Kasparov, or indeed any human brain currently playing chess." No, I don't know what an Elo rating is, but I assume I should be impressed. 

The implications in stories such as these are that we should worry that computers in league with Artificial Intelligence will relegate humanity to the sidelines of meaningful advances in, well, in what? Confronting computers in board games? Seriously now, how often do we do that? Light a fire in the hearth, open a nice bottle of brandy and invite R2D2 over for a nice game of chess?

Our paranoia regarding man versus machine is well entranced in our folklore. The ballad of John Henry tells the tale of a "steel-driving man" who successfully raced a steam-powered steel-driving engine. The victory however was short-lived as just after having been declared the winner, John Henry "laid down his hammer and he died."  Incidentally, John Henry is declared by a number of local residents to have been a real steel driver working for the Chesapeake and Ohio railroad's Big Bend tunnel in West Virginia. But that is not the point. The issue here is what do we ask of our machines, and should we be worried when they do it?

Behind a barn on my cousins Dean and Lori's farm in southeastern South Dakota is a "horsepower gear." It is a machine to which 1 to 4 horses could be hitched. The horses would walk around in a circle. The gear would translate that circular power to rotational power that could drive belts that could power a variety of other machines on the farm - conveyor belts, mill wheels, water pumps, etc. Machines like the horsepower gear shifted the need for muscle power from humans to horses and then to engines. So machines are tools designed to do the tasks that human beings don't want to do, or if we chose to do those tasks, would be an incredible waste of human muscle or brain power.

Now, there is no denying that throughout history machines have replaced human jobs. The Luddites broke up weaving mills, not because the technology was inherently bad. Actually quite the opposite, the mills were so efficient that they would replace the less efficient human weavers. Contemporary robots are replacing assembly line workers in factories like Tesla's "secret second floor" where the robots moved at such high speeds that their arms needed to be built from carbon fiber instead of steel. [Wired Backchannel 10.18.17] Robots like these will obviously reduce the number of human workers who can safely and efficiently move about a factory floor. 

The point is that we have always built machines that perform the jobs we don't want to do, or can't do simply because of our innate physical limitations.  To begin to judge our human efforts by the capabilities of our machines stands rationality on its head. I will never be able to compute the value of pi to several thousand places in less than a second. And why would I want to? Can we outrun a car, even a bicycle? Swim faster than a jet ski? Hoist more than a forklift? And let's not even think about airplanes, rockets and ocean liners. 

Similarly, why would I want to play chess or Go against a computer? If two human beings faced off across a game board and one could look up all the best moves and strategies in a huge database while the other had to relie on just their memory and instinct would we consider that match a fair assessment of their respective abilities? Of course not. But that is essentially what goes on when a human plays a computer. So by the normal rules of fair play, the computer is cheating. The results are no contest. The victories are meaningless.

And that is the point. We should not lament the fact that our machines surpass us at doing the jobs we don't want to do anyhow. It should not depress us if human beings lose games to machines designed to be "super cheaters" capable of not only stealing our playbook, but every playbook ever written.

Perhaps a better lesson to be drawn from these human versus machine events is not a consideration of how we might design ever better game-playing machines. Instead we might consider a deeper reflection on what fields of endeavor are uniquely human, beyond the ken of the most clever coder. 

I am uncomfortable with the Turing test that measures the ability of a computer program to trick human beings into believing that they are interacting with another human being and not a computer. To what end? To make our "in home personal assistants" like Siri and Alexa sound more human? Perhaps. And there may be value there, but not if we continue to treat these powerful machines like carnival attractions: Guess your weight! Tell me what cup the pea is under!

Here is a thought. Rather than trying to build a computer that can deceive our notion of humanity, why not seek to articulate notions of humanity that are utterly alien to the computational power of these machines?  Here are a few that spring to mind.

1. The soul. Despite absolutely no data to confirm the notion, most major faith-based communities espouse something like a "soul." Something that has - as of now - no discernible physical properties but is essential for a meaningful existence.
2. Life after death. Another belief with no objective data to support it that is widely spread throughout human society.
3. Love. Again we have, at best, only indirect evidence of its existence. Yet love is universally acclaimed as, if not the most powerful of human motivations, then certainly among the most powerful emotions that shape human behavior.
4. Creativity. We are constantly pushing the origins of human creativity further back in our time on the planet. While we quickly utilize every emerging technology in service to our creative impulses, the impulse to create, to express our feelings and perceptions exists independent of our machines.

There are undoubtedly more areas of human interest and concern that transcend the ability of our machines to manage or manipulate, let alone "understand" in anything like the human notion of understanding. That is not to say that we cannot harness the impressive power of our machines in service to these uniquely human arenas. But it strikes me as imperative that we need to reassess the relationship between the machines and their makers and mentors. I really don't care too much about the raw power of chess or Go playing computers. I care very much about how that power is harnessed in service to humanity. I am all in favor of the technological advances that free me from the mundane activities of everyday. So AI,  please, go ahead. Find my keys. Better yet, drive my car. Do my taxes. Wash the dishes, do the laundry. Even diagnose illness, compound medicines, perform surgery. We have human mentors who can guide you in those important tasks.

That leaves me more time to write and draw and listen to music, to go for a walk. That frees me up to reflect on all those uniquely human concerns that, I'm sorry Alexa, you just don't understand. And Siri, with all those "in service to humanity tasks" on your plate do you really have time to be playing chess?