Computer Chess

Hi, I'm Robert Lawrence with the Advantage Corporation in San Diego,

and we have a Colby 5 here,

which is highly optimized for playing chess.

And we have designed it around a recursive assembler routine

that uses a brute strength approach

to finding the optimal route.

I don't want to give away exact numbers,

but we've got the fastest depth-first search to get the most number...

We can predict more turns in advance than any other computer,

so we've got a good chance.

My field is experimental psychology,

but I've spent the last three years studying chess skill

in both humans and machines.

And this is TSAR.

This is the latest iteration of Caltech's computer chess program.

This is 3.0. Last year, 2.0 won this very same tournament.

Computers are getting smaller, they're getting better, they're getting faster.

It's a matter of time before we beat people with these things.

Do you guys have a program in the competition here today?

No. Don't know anything about it.

We're just watching 'em get ready for the end of the world here.

- World War Ill. - That's what we're here for.

Getting in on the ground floor.

Do you think a human being will ever beat a person at chess?

Between a "human being" and a "person"?

My money's on the computer.

I mean a computer...

If you ask Captain Apocalypse there... Have you talked to this guy?

I feel like I'm writing intercontinental ballistic missile routines here!

I just wanna win at chess.

What can I say, there is nothing but changes

Nothing but changes

What can I say, there is nothing but changes

Nothing but changes

You and I stay the same anyway

Does it really take time to be free of your mind?

Hey! Hey!

Do not ever shoot at the sun, you're gonna burn out the tube!

- I wasn't shooting at the sun, it's fine. - You're shooting at the sun!

You're gonna shoot the rest of the weekend inside.

We want to be only inside. That's it!

But computer chess goes back even further,

to the mechanical Turk, 1770, the original chess-playing machine.

It played against and defeated Benjamin Franklin

and Napoleon Bonaparte.

Well, he had a secret.

There was a human being inside the machine.

Though the first was a fraud, we're working to get back

to that level of chess play when machine beats man.

I greet you for our annual North American Computer Chess Tournament

and present you with a panel of the best and the brightest.

From my left, there is Les Carbray from Allied Laboratories.

Hi there.

Martin Beuscher, sitting in for Tom Schoesser.

And that's last year's winner, TSAR 2.0 from Caltech.

Roland McVey from MIT, the programmers of STASIA.

And finally Mike Papageorge, who is an independent programmer.

Is there a computer program in the house

which can stand up to a human chess master?

That's me! When will a machine beat me?

Many years ago I made a public wager

that no computer would beat me until the year 1984,

and that date is quickly coming up.

So what do you think, gentlemen? Will I win my bet?

I... I think you're cutting it close,

but I think you will win that bet.

If I was in your shoes, I wouldn't extend it any.

I think within two years of that, 1986,

you won't stand a chance,

and I think that, say another ten years after that,

there isn't a man or woman alive who will stand a chance.

Well, there may be some disagreement,

but look at the speed with which the technology, the software,

but particularly the hardware, is progressing right now.

And just based on that speed,

barring a calamitous event like World War Ill to set us back,

I truly believe it to be inevitable.

How about any of the other gentlemen, do you have a thought on this?

Want to chime in?

This articulate panel?

We've experimented some with parallel processing

and then we're doing selective search,

and all the algorithmic and software things that we're doing

makes a difference, it does make an improvement,

but it's dwarfed by the improvement we get just with better hardware.

More memory and faster processor.

We're blessed at Allied in that we have the equipment,

we have the resources at our disposal, so...

The person I'm really excited to talk to about real gains in software

is perhaps Tom Schoesser,

who I thought was going to be here on this panel?

Yes, Professor Schoesser will be here shortly,

he is delayed, but he will be here.

OK, could you perhaps talk about some of that?

Talk about advances? Well, I...

I think it's safe to say that we're operating in some of the same avenues as Allied,

but maybe Professor Schoesser could speak more to that,

but I'm not necessarily certain he would speak about it.

Well, anyway, whatever you did last year was working

because TSAR 2.0 was the winner,

and they were especially good about endgames,

and that's a place where computer programs often falter.

So I think what I'd like to show is an example of what happened last year.

And I'm sorry to do this to my friend right here, but...

So, STASIA versus DAWN.

Well, this should have been an easy win for STASIA.

Any human being can see that there's a forced queen-takes-queen,

and it's a very easy victory for White.

But instead... Queen checks.

Queen checks.

Check, check, check, check, back and forth,

over and over again, lost in a loop.

Very, very embarrassing, wouldn't you say?

Yes. It was not our finest moment.

But you bring up a good point.

I think we've seen a lot of increase in hardware speed,

it's definitely led to stronger game play.

In fact I think we're searching the tree deeply enough now

that we're catching pretty much all of the tactical issues.

But STASIA's greatest weakness,

which I think is the same weakness all our programs share,

is that it really has a very poor understanding of the positional issues.

So this year at MIT we've hired a grandmaster to help us out.

He's helped us come up with a couple rules on the endgame specifically

that will hopefully help us avoid the comedy that we saw last year.

Mr Papageorge, we haven't heard from you yet.

You usually have a very unique opinion, so...

Frankly Pat,

I have to tell you I find the programming of my fellow competitors here

to be almost as boring as this discussion.

Why, because the machine can't compete against the human soul?

Very interesting!

Sure, sooner or later somebody's gonna write a program

that's gonna beat you in '83, in '85, who cares?

Listen, you guys are just trying to eke out

one little victory versus each other, you know?

You get here this year, you get here next year.

My program is seeking harmonies, seeking innovations...

- I'm not entirely sure... - That's what I'm doing!

From the perspective of programming that it actually means anything.

Well, we'll see when we get to the competition.

Could you please spell your name for me, sir?

Sure, it's P-A-P-A-G-E-O-R-G-E.

I don't see anything here.

- I'm sorry. - P-A-P-A-G-E-O-R-G-E?

- No. - M-I-C-H-A-E-L?

I'm sorry.

I could call to one of the neighboring hotels and see if there's an available room.

No, I need to stay at this hotel where the competition's being held!

I am Pat Henderson, I am very excited.

We have players from all over America, Canada, Quebec...

Just great, and we have something new this year,

we have a lady who is competing, way in the back corner.

I'm happy about that too, she's welcome.

So, the way this tournament works is this,

we play Swiss system - that's five rounds.

That means everybody keeps playing. If you lose, you play to the end.

However there is one big winner,

and that winner gets a $7,500 prize.

And the other thing you get to do, you get to play me.

Stick around for the last day

because I challenge that winning chess program.

This will be a grueling couple of days, so let's keep it light,

let's keep it collegial, be respectful, and of course have fun.

Let's begin round 1.

So, Tom Schoesser - is he showing up?

He's a good guy. Way ahead of his time.

He's like Tesla, he's so far ahead.

I wanna talk to him about the "Eureka moment".

And frankly, we're on the way to a kind of timid strategy on the side of White.

I think the reason is, they are playing the very, very strong Caltech TSAR 3.0 team.

In this case, maybe the program is trying for a draw?

In fact I defy anybody to tell me any theory of Tesla's

that's been disproven today.

I mean, wireless energy transfer,

radio before Marconi,

the spark plug,

alternating current engine...

Rook to D8.

- Maybe the Industrial Revolution. - No, it says E8...

TSAR takes pawn.

Queen takes queen.

And rather than capture queen, the rook just approaches it.

Is there any possibility that this is...

some kind of very advanced...

I mean...

we just gave up a queen to...

to get his queen there and now we're basically just...

Well...

- Don't move that rook. - All right?

- Be right back. - Excuse me! Excuse me...

There seems to be a little glitch here,

maybe we should just follow with the camera

and I'll look over here at the STASIA SHORTSLEEV game.

We're having serious technical problems,

- I need to run a very, very brief diagnostic. - I'm sorry, once the clock has started...

There's a tiny... I'm almost certain I know what it is,

it's just going to take a moment of our time.

I appreciate that,

but unfortunately once the clock has started we cannot halt the game.