F1 To Be Decided By Wins

[Grabthaw the Hammerslayer 4]

The AI idea is not a terrific one - the problem is that using standard AI you would end up with a moving road block taking all the corners perfectly....

One of the things that makes a human "human" is their flaws, personality, etc. Hard to replicate, especially the changes and fluxes - e.g. someone carves you up, you react aggressively, etc... The team boss comes on the radio and tells you to slow down to help your team mate, you sulk, you get a high from passing someone and this fires you on to pass someone else.

Moreover what's the point?

And besides, it would be impossible to replicate Alonso's eyebrows by any artificial means

[Jean Todt 4]

Bernie wants this idea to SAVE him paying for those lesser teams that currently score points and get paid for transportation costs etc, because of the points they have won.

The GREEDY OLD b#####d is just looking after the pot of gold that he collects from every race. F(*&^%$#%^&* whatever happens to the SPORT of F1

The sooner Bernie is off and out of F! and for that matter the surface of Earth the better F1 will be.

Amen.

[maure 1]

OK so we both agree that they probably could be developed but you object to me saying it would likely be easy. Well, I think you'll find the people involved are merely scientists and therefore ''prostitutes'' (in Maure's ''fantasies'') who will probably be the first to admit they are ''easy''.

Your inference about everyone's understanding of a topic that they have corrected you on says a lot about your appreciation of their expertise. Please, do a literature search for all the seminal contributions I myself have made to the field of pattern recognition, and other cognate disciplines.

Drivers would be left trailing behind anyway. It would be as safe as it is now.

Innovations that already EXIST aren't innovation any more. Once upon a time, using a wheel was enough to make a caveman cool but now even Force India understand the basics of wheel technology. But I know that doesn't bother you, Maure: given your lack of expertise in such technology I suppose you can sit back and enjoy F1's being the pinnacle of wheeliness.

And so, unable to pursue any further the exvcuse that I am against AI's, you change your tune to... what again? Your claim to have made "seminal" contributions to what "cognate disciplines"? As expected as it is hilarious.

Beyond this latest goof up, you post some additional handwaving concerning regulations that prohibit existing techonology although, you claim, allow its innovation. Not as funny as your "academic credentials" but worthy of a chuckle.

Now. Please post here a list of references detailing your academic work on the subject of pattern recognition. Don't run around. Don't pretend it away. Don't post a bigger lie in another futile attempt to cover your last one. Post the references to your work. Let's review it together...

[maure 1]

The AI idea is not a terrific one - the problem is that using standard AI you would end up with a moving road block taking all the corners perfectly....

One of the things that makes a human "human" is their flaws, personality, etc. Hard to replicate, especially the changes and fluxes - e.g. someone carves you up, you react aggressively, etc... The team boss comes on the radio and tells you to slow down to help your team mate, you sulk, you get a high from passing someone and this fires you on to pass someone else.

Moreover what's the point?

And besides, it would be impossible to replicate Alonso's eyebrows by any artificial means

Actually, the subject has some interest and work has been done on it, inside and outside the track. We are just too far for it to matter. Of course, some contributions in that direction have been made like launch and traction control, etc. But the jump to a full fledged autonomous AI driving an F1 (over the tracks we've got nowasdays) is not likely to happen any time soon. Simulations, sure. Actual AI's able to perform like human drivers, not quite yet.

[cavallino 2]

So, first and _one_more_time_, that it might be possible to develop those AI's _does_not_imply_ that they exist today or that their development is "easy" as you, my loopy fellow, claim.

There are already cars that drive themselves. As long as we can make the car know its exact position and orientation on the track (easy enough) I am sure we can make cars that can do test laps pretty fast. While judging overtaking etc I would think humans would stll have an edge.

Second, your equating chess playing with F1 racing says a lot about your poor understanding of the subject. Entertain yourself doing a google search (which is, after all, what _you_ consider "reputable" research) and find out the important differences that exist between the two problem domains. You can start anywhere, say, on pattern recognition...

The pattern recognition part isn't that hard, unless you take out positioning systems. Even then..

Third, your desire to see AI's racing humans drivers is _again_ indication that you have no clue regarding the problem domain. Heck, even from the point of view of driver safety, it would be unsustainable.

You are clueless. AI drivers will be a lot safer than human drivers.

And so, unable to pursue any further the exvcuse that I am against AI's, you change your tune to... what again? Your claim to have made "seminal" contributions to what "cognate disciplines"? As expected as it is hilarious.

I don't have reason to doubt his claims. However most of your posts are arguments from authority which make very little sense without presenting any credentials. I know more than enough about the fields in question, I don't remember MW ever making specious claims though I have my disagreements with him of course. Most of what you post is a load of crap, so unless you can back it up, it merely funny.

Beyond this latest goof up, you post some additional handwaving concerning regulations that prohibit existing techonology although, you claim, allow its innovation. Not as funny as your "academic credentials" but worthy of a chuckle.

Now. Please post here a list of references detailing your academic work on the subject of pattern recognition. Don't run around. Don't pretend it away. Don't post a bigger lie in another futile attempt to cover your last one. Post the references to your work. Let's review it together...

How about you post yours, I am more interested in those. All the claims you make here and the ones you made about climate change ?

[dribbler 6]

There are already cars that drive themselves. As long as we can make the car know its exact position and orientation on the track (easy enough) I am sure we can make cars that can do test laps pretty fast.While judging overtaking etc I would think humans would stll have an edge.

But not over Coulthard.

[maure 1]

Since Murray has just claimed that he is published "seminal contributions" on the subject, cavallino jumps in to help him out... with his own brand of goof ups.

There are already cars that drive themselves. As long as we can make the car know its exact position and orientation on the track (easy enough) I am sure we can make cars that can do test laps pretty fast. While judging overtaking etc I would think humans would stll have an edge.

If you were _truly_ familiar with autonomous intelligent systems, you would know that it is far from "easy enough" to have an F1 racing around a track by simply "judging" its orientation. In fact, you yourself (and naively) put forth the problem of overtaking... how? Aren't you just going to say that pattern recognition "isn't that hard"?

It might just be you don't even know what it is you are not even aware of what you are saying... yourself.

The pattern recognition part isn't that hard, unless you take out positioning systems. Even then..

Not that hard. Really. The way you guys fantasize about science never ceases to amuse me.

You are clueless. AI drivers will be a lot safer than human drivers.

Will? In what future WILL that be so?

I don't have reason to doubt his claims. However most of your posts are arguments from authority which make very little sense without presenting any credentials. I know more than enough about the fields in question, I don't remember MW ever making specious claims though I have my disagreements with him of course. Most of what you post is a load of crap, so unless you can back it up, it merely funny.

Funny, indeed, that all the evidence to support your arguments is... missing.

How about you post yours, I am more interested in those. All the claims you make here and the ones you made about climate change ?

It was Murray who claimed to have published on the subject. That you don't ask HIM to back his own words with evidence says everything... and that you tiptoe around his claim that AI's for chess have anything to do with F1 says even more... about your own knowledge of the subject.

But it is all to be expected. The surprise would've been that you actually knew something about the topic and that your post was not another desperate run around in order to accomplish... what exactly? A personal attack? All you've accomplished in that regard has been to amuse me with your desire to parade your ignorance on yet another subject.

Let's think about a future when you have something to say beyond "it is all that easy" in regards to problems of considerable difficulty...

Freaking jokers.

[maure 1]

It is still funny. Thank you for this.

Forget your lack of knowledge on the subject, people. You guys don't even have the common sense of saying something like "... it would be an extremely complex problem, but I think it could be done...". No, no. For you guys, it is all easy, trivial. Been there, done that. No problem at all...

Hilarious.

I mean, Murray, come on, man! Don't you even realize that by claiming to have made "seminal contributions" in pattern recognition and "cognate disciplines" you are claiming to be an expert on several fields? There are fellow academics out there putting 70 hour weeks to simply be _relevant_ on their field and you run your mouth about being an expert on several fields.

Are there truly no limits to your delusion... honestly.

[cavallino 2]

If you were _truly_ familiar with autonomous intelligent systems, you would know that it is far from "easy enough" to have an F1 racing around a track by simply "judging" its orientation.

Orientation and exact track position from positioning systems/ gyros. Velocity through - oh boy that's a hard one wonder how we can measure velocity Sensors on the tyres in addition to all the stuff they use for traction control. And then test and optimize algorithms like mad, use learning. Sounds pretty doable I would say.

In fact, you yourself (and naively) put forth the problem of overtaking... how? Aren't you just going to say that pattern recognition "isn't that hard"?

Depends on how much you let the computers cheat. If you let them know the position etc of all other cars, it becomes much simpler.

It might just be you don't even know what it is you are not even aware of what you are saying... yourself.

I am a computer scientist with more than a passing interest in AI - unlike the climate change thread, I don't even need to google for this..

Will? In what future WILL that be so?

AI drivers can respond to loss of traction etc far faster than any human, there is a reason traction control worked so well.

It was Murray who claimed to have published on the subject. That you don't ask HIM to back his own words with evidence says everything...

Well if he was talking crap and refused to back it up...

and that you tiptoe around his claim that AI's for chess have anything to do with F1 says even more... about your own knowledge of the subject.

He didn't compared the problems, merely made a statement on how computationally demanding it would be. He is right, I am sure a supercomputer that weighs the same as a driver will be able to cope.

Let's think about a future when you have something to say beyond "it is all that easy" in regards to problems of considerable difficulty...

I am sure it's a problem the teams can take on with their huge budgets. In universities we tackle pretty sophisticated and far more open ended problems with a fraction of the budget. As I said, overtaking would be the tricky bit but then there isn't any overtaking in f1 anyway

[Quiet One 15]

The central argument (medals) is so idiotic that it's not even worth discussing. As the thread seems to have degenerated into some argumentative dead end over AI cars, here is my contribution (as usual, more out of mere arrogance than from actual knowledge, I know nothing about electronics or cars, for that matter)

1) AI cars should be easy. If you can program them in a video game, you can certainly make a similar program for actual cars. Just throw in some servos and that stuff that Mythbusters use to remote control cars and voila.

2) Cars can be kept inside the track via means of buried wires or sensors. If the car crosses the line it gets disconnected. If you are going to say that that would not be like nowadays, when cars can go off trtack and then return, I will just reply: AI cars will be nothing like nowadays!

3) Now, you can argue with me. At least I am the only one who actually provided a theory on how would AI cars be. It might be idiotic,. but it will be much more fun than the merely dialectic pyrothecnics you guys are engaged in

4) One final word: AI cars will be something so much more idiotic than medals it isn't worth the effort either. But it's the off-season...

[maure 1]

Orientation and exact track position from positioning systems/ gyros. Velocity through - oh boy that's a hard one wonder how we can measure velocity Sensors on the tyres in addition to all the stuff they use for traction control. And then test and optimize algorithms like mad, use learning. Sounds pretty doable I would say.

Depends on how much you let the computers cheat. If you let them know the position etc of all other cars, it becomes much simpler.

I am a computer scientist with more than a passing interest in AI - unlike the climate change thread, I don't even need to google for this..

AI drivers can respond to loss of traction etc far faster than any human, there is a reason traction control worked so well.

Well if he was talking crap and refused to back it up...

He didn't compared the problems, merely made a statement on how computationally demanding it would be. He is right, I am sure a supercomputer that weighs the same as a driver will be able to cope.

I am sure it's a problem the teams can take on with their huge budgets. In universities we tackle pretty sophisticated and far more open ended problems with a fraction of the budget. As I said, overtaking would be the tricky bit but then there isn't any overtaking in f1 anyway

So you continue to trivialize the problem while claiming to be a "computer scientist" working at a university... your conclusion? While budgets in the millions have been spent (and continue to be) on the subject, our resident genius, cavallino, claims the matter "sounds pretty doable".

How can I not laugh?

[maure 1]

The central argument (medals) is so idiotic that it's not even worth discussing. As the thread seems to have degenerated into some argumentative dead end over AI cars, here is my contribution (as usual, more out of mere arrogance than from actual knowledge, I know nothing about electronics or cars, for that matter)

1) AI cars should be easy. If you can program them in a video game, you can certainly make a similar program for actual cars. Just throw in some servos and that stuff that Mythbusters use to remote control cars and voila.

2) Cars can be kept inside the track via means of buried wires or sensors. If the car crosses the line it gets disconnected. If you are going to say that that would not be like nowadays, when cars can go off trtack and then return, I will just reply: AI cars will be nothing like nowadays!

3) Now, you can argue with me. At least I am the only one who actually provided a theory on how would AI cars be. It might be idiotic,. but it will be much more fun than the merely dialectic pyrothecnics you guys are engaged in

4) One final word: AI cars will be something so much more idiotic than medals it isn't worth the effort either. But it's the off-season...

Don't be surprised if I tell you that yours is the most reasonable post on the subject so far.

Indeed, the first problem with these goofs is that they confuse simulation with reality (at best... and I'm being very very generous here).

The buried wires solution is not mandatory and never ideal. A car is not a train. Fully autonomous driverless vehicles should be able to make sense of their surroundings. These has been accomplished already to some extent... see the DARPA Grand Challenge, for example, and note the average speeds achieved.

[Quiet One 15]

Don't be surprised if I tell you that yours is the most reasonable post on the subject so far.

Of course. I am not surprised at all

The buried wires solution is not mandatory and never ideal. A car is not a train. Fully autonomous driverless vehicles should be able to make sense of their surroundings. These has been accomplished already to some extent... see the DARPA Grand Challenge, for example, and note the average speeds achieved.

I guess a car going at 300+ km/h cannot depend on buried wires (I am not even sure how do those wires work, anyways) I don't think that following some kind of buried guide would work, and it would be too boring. But electronical "boundaries" can be marked by means of sensors at the sides of the track and in the cars.

Again...what would all that be for? I don't wanna see a bunch of automated cars running. I'd rather watch Lewis crashing onto kimi's car anyday.

[freaky2 1]

Of course. I am not surprised at all

I guess a car going at 300+ km/h cannot depend on buried wires (I am not even sure how do those wires work, anyways) I don't think that following some kind of buried guide would work, and it would be too boring. But electronical "boundaries" can be marked by means of sensors at the sides of the track and in the cars.

Again...what would all that be for? I don't wanna see a bunch of automated cars running. I'd rather watch Lewis crashing onto kimi's car anyday.

As long as there are people crashing into Lewis' car as well...

(Yes, yes, another very intelligent post from me...... xD)

[Max Mosley 2]

No, no. For you guys, it is all easy, trivial. Been there, done that. No problem at all...

Yes most problems in F1 do fall into the easy category, for experts. I've noticed people whose expertise is in other areas or none do find that surprising. I try to educate you all but we have some slow learners. Bless.

I mean, Murray, come on, man! Don't you even realize that by claiming to have made "seminal contributions" in pattern recognition and "cognate disciplines" you are claiming to be an expert on several fields? There are fellow academics out there putting 70 hour weeks to simply be _relevant_ on their field and you run your mouth about being an expert on several fields.

I'm glad you said that. The colleagues of mine that earn my deepest respect are those for whom science doesn't come so easily. If there were any justice in the world they would be the ones publishing the seminal works in the field.

I am a computer scientist with more than a passing interest in AI - unlike the climate change thread, I don't even need to google for this..

Cav and I send each other our latest journal articles in these fields, you see. If you have some any interesting research of your own to contribute, feel free to join the club.

It might be idiotic,. but it will be much more fun than the merely dialectic pyrothecnics you guys are engaged in :

You're trying to wind us up!

[maure 1]

Of course. I am not surprised at all

I guess a car going at 300+ km/h cannot depend on buried wires (I am not even sure how do those wires work, anyways) I don't think that following some kind of buried guide would work, and it would be too boring. But electronical "boundaries" can be marked by means of sensors at the sides of the track and in the cars.

Again...what would all that be for? I don't wanna see a bunch of automated cars running. I'd rather watch Lewis crashing onto kimi's car anyday.

It has to work on visual feedback, radar, lasers, infrared, etc (not wires) because it is a race what we are talking about. This has been somewhat addressed already for street vehicles. But there is an even bigger problem. It is referred to as "motion planning" (or as the "navigation problem") and this one is still not figured out. Considerations regarding racing without a satisfactory solution to this issue are pointless.

Of course, there is also the safety issue (despite cavallino's dellusions on the subject, present or future). During a race, mechanical failures are not uncommon. If there were a malfunction of any of the electronics running the AI, the car would become a missile. Not only any sane human driver would refuse to be on track against such potential lethal weapon, even spectators would be at risk. And no, fellow-poster fantasies regarding remote kill-switches are no guarantee.

That said, is a driverless F1 an interesting proposition? Yep. As a parallel competition. The technical challenge alone is worth the effort but not worth the substitution of the present competition. It's the same with all sports. That bots can do certain things (better or worse) is no reason to pursue that venue _only_.

[maure 1]

Yes most problems in F1 do fall into the easy category, for experts. I've noticed people whose expertise is in other areas or none do find that surprising. I try to educate you all but we have some slow learners. Bless.

I'm glad you said that. The colleagues of mine that earn my deepest respect are those for whom science doesn't come so easily. If there were any justice in the world they would be the ones publishing the seminal works in the field.

Cav and I send each other our latest journal articles in these fields, you see. If you have some any interesting research of your own to contribute, feel free to join the club.

You're trying to wind us up!

Your wounds are self-inflicted, my friend.

[Quiet One 15]

It has to work on visual feedback, radar, lasers, infrared, etc (not wires) because it is a race what we are talking about. This has been somewhat addressed already for street vehicles. But there is an even bigger problem. It is referred to as "motion planning" (or as the "navigation problem") and this one is still not figured out. Considerations regarding racing without a satisfactory solution to this issue are pointless.

Ok, I see the point. I was talking about merely making some cars race with no driver. That can't be so much harder than a slot car race (and almost as entertaining). You are talking about something more ambitious. Well, if they ever decided to go the AI way, I bet they will choose my "dumb cars running in circles" path over your "cars with no driver acting almost human".

But yes, for cars to act with such a degree of precision, you would need an almost human brain. If they can't solve the "true vision" AI problem, I don't think we will see anything too complex anytime soon.

Of course, there is also the safety issue (despite cavallino's dellusions on the subject, present or future). During a race, mechanical failures are not uncommon. If there were a malfunction of any of the electronics running the AI, the car would become a missile. Not only any sane human driver would refuse to be on track against such potential lethal weapon, even spectators would be at risk. And no, fellow-poster fantasies regarding remote kill-switches are no guarantee.

Yes, if the car malfunctions, it will become a missile. Much like today, only it won't carry a likely victim inside. Adequate safety measures would not be more complex than today. They should be easier, if any. Bigger runoff areas, more space between the public and the cars and that's it. No ambulances, no medics.

Kubica's crash last year was a great example of a car at maximum speed totally out of control. The only persons in danger were the drivers and the track personnel (if there were some of them around). It wasn't too bad. I can't see why it would be worse without humans involved.

That said, is a driverless F1 an interesting proposition? Yep. As a parallel competition. The technical challenge alone is worth the effort but not worth the substitution of the present competition. It's the same with all sports. That bots can do certain things (better or worse) is no reason to pursue that venue _only_.

As a parallel competition? Mmnmmh...I agree. They are trying to make some robots play football, after all. Still, my own preferences go towards driver controlled cars. But yes, it would be interesting to watch them try!

[Jez 0]

Moose 11, I think you can say goodbye to your thread.

[cavallino 2]

The central argument (medals) is so idiotic that it's not even worth discussing. As the thread seems to have degenerated into some argumentative dead end over AI cars,

Degenerated? I prefer the term 'blossomed'

1) AI cars should be easy. If you can program them in a video game, you can certainly make a similar program for actual cars. Just throw in some servos and that stuff that Mythbusters use to remote control cars and voila.

Well not quite. Computer AIs cheat most of the time and the simulation is pretty basic.

2) Cars can be kept inside the track via means of buried wires or sensors.

We can do better than buried wires and sensors now..

So you continue to trivialize the problem while claiming to be a "computer scientist" working at a university... your conclusion? While budgets in the millions have been spent (and continue to be) on the subject, our resident genius, cavallino, claims the matter "sounds pretty doable".

Budgets of millions have been spent on designing AI racecars?? Where?

The buried wires solution is not mandatory and never ideal. A car is not a train. Fully autonomous driverless vehicles should be able to make sense of their surroundings.

They dont need to make sense of their surroundings. They are driving in a small well defined environment. Who cares about the surroundings, as long as the car knows where it is on track.

These has been accomplished already to some extent... see the DARPA Grand Challenge, for example, and note the average speeds achieved.

bulls##t. The DGC is not relevant here, it has cars negotiating traffic rules, urban environments etc. The problem of driving an F1 car around a track is much simpler.

[cavallino 2]

I guess a car going at 300+ km/h cannot depend on buried wires (I am not even sure how do those wires work, anyways) I don't think that following some kind of buried guide would work, and it would be too boring. But electronical "boundaries" can be marked by means of sensors at the sides of the track and in the cars.

You don't need wires. It is perfectly possible for a car to know exactly where it is on track using positional systems. Mauron son of Sauron would have told you this if he had a clue.

It has to work on visual feedback, radar, lasers, infrared, etc (not wires) because it is a race what we are talking about. This has been somewhat addressed already for street vehicles. But there is an even bigger problem. It is referred to as "motion planning" (or as the "navigation problem") and this one is still not figured out. Considerations regarding racing without a satisfactory solution to this issue are pointless.

Navigation? Motion planning? FFS stop googling to pretend you have a clue. It is a one way round track. You need n navigation, or motion planning. A rudimentary 'don't crash into cars around you' algorithm would be a great start.

During a race, mechanical failures are not uncommon. If there were a malfunction of any of the electronics running the AI, the car would become a missile.

You're equating mechanical failures with electronic failures? Are you aware of the amount of computing power involved in driving airplanes for example?

But yes, for cars to act with such a degree of precision, you would need an almost human brain.

No you don't. Its a limited well bounded problem.

Kubica's crash last year was a great example of a car at maximum speed totally out of control.

Indeed. An AI driver would have caught the signs of a car going out of control far earlier and responded instantaneously. No human can compete with that.

[Medilloni 6]

So would the medals go to the programmers?

Would the programmers go for bust if they're not in a medal position?

Will the cars have a finger it can raise to the other fingers in the other cars computers?

Will team orders be allowed if one program is not as fast as another, and will the loooooooser program sulk?

Will anyone watch this sh!t?

[monza gorilla 1]

Does anyone care about 3 guys having a p!ssing contest?

[Quiet One 15]

Degenerated? I prefer the term 'blossomed'

You should have been a scholastic Christian in some previous re incarnation...

Well not quite. Computer AIs cheat most of the time and the simulation is pretty basic.

As you say somewhere else, it can't be that hard to make the cars go around the track without bumping each other. And that's what video games AI cars do, even the more basic ones.

We can do better than buried wires and sensors now..

Well, like I said, my knowledge is next to zero in these matters.

They dont need to make sense of their surroundings. They are driving in a small well defined environment. Who cares about the surroundings, as long as the car knows where it is on track.

Maure talks about cars that could behave as if human drivers were controlling them. I don't think it would make much of a difference, from the outside.

You guys are discussing two different views on AI cars. Maure's is more complex, Cav's (and mine) much simpler. I can see the challenge in Maure's approach, I am not so sure it would be so different as throwing some cars in the track with a race track map loaded and some proximity sensors. But it will certainly be a challenge.

It's like chess computers. The old approach was just making it analyze every combination possible. Simple. Nowadays they are trying to make it more "rational".

[Fed up 0]