maandag 18 september 2017

Turing’s Duck

In reply to those convinced that “Robots will Never…”

24/7 life-like performance of waggle and quack

In 1950 Alan Turing wrote his paper “The Imitation Game”. The paper is still relevant and keeps providing food for thought, well beyond the popular Turing Test(1) it proposes.

In this paper Alan addresses the question “Can machines think?” and defends his own position ("Kind of Yes") well against the many counter-arguments of the time. So much so that even modern day nay-say-article-authors often fail to show they used the advantage of the past 65+ years to come up with new or stronger arguments.

Before entering this arena or contemplating your own strategy for attack I would recommend adversaries to fully appreciate the triple-F characteristics of the elegant weapon of choice Turing crafted to defend his case: it is Falsifiable, Factors in the observer and is ultimately Fair.

Falsifiable

Whatever your position is, you don't control the future. So avoid being the next guy topping the charts of "worst tech prediction" by simply not declaring absolute knowledge about anything in absolute time (like 'never').

Being a stated absolute take-it-or-leave-it fact, there is just no more conversation to be had(2). When the prophets get caught up by the facts there is the classical two-way option to reduce the cognitive dissonance: admit to the new facts and learn, or stubbornly ignore: i.e. say people didn't understand what you meant, focus on whatever remaining gap, invent new thresholds and rules, etc.

In fact there is the third option: being about the unreachable eternal "ever", there is a fair chance you don't live long enough to experience the confrontation. Or claim your prize. So what?

Staying relevant well after your expiry date requires a more subtle approach. Turing shows what correct scientific 'falsifiable' argumentation looks like: Do not just claim (that never...), instead provide a practical test (how would we know?).

What follows is an interesting dialogue, a negotiation about how high the bar should be set, honest reflections on what we currently do ourselves not being all that perfect either. It spawns an industry and future research questions. A field populated by those searching to close the gap, committed to ever do better.

In short: it lights a path to progress. A smart angle to sail up against the wind of "never possible" defeatism. Honestly, you 're only baiting us.

Factors in the observer

The Turing Test is not (only) about the robot. And it is most striking how much people ignore that fact. The imitation game surely puts a challenge to the imitator (the robot) to wash away any differences that blow their cover. Finally though it is the 3rd party in the game (the judge) who is calling the shots: whenever he can't see the difference the imitation actually is the real thing.

This "cannot see" comes in many forms: running out of smart questions to ask, or simply failing to care any further. These are all versions of "accepting it is good enough". Turing might have been a trained mathematician, he surely has all the looks of an engineer as well.

There is a special brand of "not in my domain it will not ever" peeps. To them, there is this version. Yo, Ivory Tower Expert with the exquisite emolument for exceptional services: "Thou might know best thou areth the smartest amongst thou peers." But the humble man in dire need of your wisdom doesn't. And that might be both your (current) strength and your (future) weakness. For that customer "good enough at the right price" will nicely do, thank you. They decide who your competition really is.

If you fail to see this simple truth, you will not even see it coming.

Fair

Fitting like the blindfold on Lady Justice, there is this "Veil of Ignorance" separating the judge from the players in the Imitation Game. Each side gets an equal chance in this game. And it makes Turing's approach rather provocative. By design it leaves the possibility open.

Unlike the nay-say-author that benefits from his upfront knowledge to label us humans versus them bots in his article to eagerly protect the "Carbon Privilege"; the Judge does not have the luxury of escaping into the semantic tricks we are so used to that we even don't notice we are doing it. Upfront the word "thinking" is one we have reserved for that activity only we humans do. No matter how fast the submarine, it may dive deeper or float better, it will never "swim". The fact that airplanes actually do "fly" only reveals we don't have wings ourselves.

But Alan's setup makes it impossible to cheat.

Imitating is not cheating

I've come to believe the "The Imitation Game" is a metaphor for the life we lead. Honestly. Biologically life’s essence is reproduction (copy with variation) but socially it isn’t that much different: We grow up copying each other. The game of ‘fitting in’ requires this growing perfection of mimicking our peers. Through formal education we believe we ‘become’ a lawyer, surgeon, engineer, while in reality we are merely training our ability to convincingly role-play that part in society.

To be, as far as we know, is to imitate. Desperately. Since it isn't cheating, you may even take pride.

-oOo-

Closing remark

The Loebner Prize is an annual actual "Turing Test" competition measuring up the state of affairs of AI progress in this imitation game. Interestingly enough, there is a side-bet going on for the human players in the game as well. The price for the most human human. The person that was most easily spotted as that (faulty, messy, emotional, needlessly repetitive, whimsical, clumsy, … never dull) actual person.

See this account(3) of the interesting 2009 winner (The video takes only 4')



So gradually, trying to convince the judges, even as a human, becomes an art in itself.

As robots become the new normal, I see lots of humans eagerly lifting themselves over some magical threshold into this new privileged pack of ‘exceptional beings’. It makes me wonder which of the two groups will eventually match the semantics of the word “artificial” best.

-oOo-

This article is part of a series: Turing's mirror - Turing's Duck - Turing's Razor

-oOo-


footnotes

(1) Note that, similar to his 'machine' it was only after time that this 'game' was named after its inventor and became known as 'the test'.

(2) This "take it or leave it" approach is the same for your religion, other elements of your identity, or sadly almost any arguable position in contemporary social issues. Somehow we forgot how to state 'negotiable' positions and are thus closing the door to any compromises, balanced solutions or win-wins. We need to learn how to recognize this kind of tactics upfront and simply refuse being fooled into negotiations with people entering the discussion like this. If reasonable people are to prevail they have to learn to be absolutely unreasonable to people showing lack of (willingness to) reason. Hell, maybe bots could help out!

(3) If you have more time to spend, I can warmly recommend this more elaborate Radiolab Podcast: Talking to machines.

Geen opmerkingen:

Een reactie posten