FIA Thread

[MadMax]

I agree. Most sports need to evolve faster in adoption of technology and removing human error element.

Anything designed and made by humans, even AI, will have human failings baked in to it. Humans can't make perfect systems.

And yet, the world is running on human built systems! What's more, we have even reached Mars with such imperfect systems!

The world goes wrong quite often. You might see it in your favorite evening news bulletin.

Mars missions have something like a 70% failure rate, either in the launch phase (rocket fails), the cruise phase, or the landing (usually in the entry and descent phase). Some even get to the surface safely and then fail to deploy successfully.

Any AI system is only as good as the programming and the teaching it's given. Both of those are prone to human bias and error. And if you use an AI to design and teach another AI, you'll still have the original errors and biases baked in to the original and these will be inherited to some degree by subsequent systems.

Humans aren't God. We can't do perfect.

[mendis]

Anything designed and made by humans, even AI, will have human failings baked in to it. Humans can't make perfect systems.

And yet, the world is running on human built systems! What's more, we have even reached Mars with such imperfect systems!

The world goes wrong quite often. You might see it in your favorite evening news bulletin.

Mars missions have something like a 70% failure rate, either in the launch phase (rocket fails), the cruise phase, or the landing (usually in the entry and descent phase). Some even get to the surface safely and then fail to deploy successfully.

Any AI system is only as good as the programming and the teaching it's given. Both of those are prone to human bias and error. And if you use an AI to design and teach another AI, you'll still have the original errors and biases baked in to the original and these will be inherited to some degree by subsequent systems.

Humans aren't God. We can't do perfect.

So stop trying? I understand why you think so as soon as you said God. You do realize that you are using a complex technology to communicate with me here right? Something where a number of highly motivated people that have tried, failed and succeeded over decades to enable this. This sports itself involves blood and sweat of engineers that have enabled competing machines using ultra complex technologies.

If you do not try, you don't belong.

[MadMax]

And yet, the world is running on human built systems! What's more, we have even reached Mars with such imperfect systems!

The world goes wrong quite often. You might see it in your favorite evening news bulletin.

Mars missions have something like a 70% failure rate, either in the launch phase (rocket fails), the cruise phase, or the landing (usually in the entry and descent phase). Some even get to the surface safely and then fail to deploy successfully.

Any AI system is only as good as the programming and the teaching it's given. Both of those are prone to human bias and error. And if you use an AI to design and teach another AI, you'll still have the original errors and biases baked in to the original and these will be inherited to some degree by subsequent systems.

Humans aren't God. We can't do perfect.

So stop trying? I understand why you think so as soon as you said God. You do realize that you are using a complex technology to communicate with me here right? Something where a number of highly motivated people that have tried, failed and succeeded over decades to enable this. This sports itself involves blood and sweat of engineers that have enabled competing machines using ultra complex technologies.

If you do not try, you don't belong.

Oh, I'm not religious at all. It was an attempt to demonstrate the difference between where we are and where would need to be to make things perfectly. Just about every culture has some form of deity that is perfect and thus it's a simple cross-cultural reference point.

As to the rest of your post, yes, computers are amazingly complex and technical things. And they go wrong, even though many thousands of very clever people have designed, built and programmed them.

My original point was that even if we had an AI system checking for off track excursions, etc., as suggested earlier in the thread, it would still be liable to biases and errors just as the current human adjudicators are - because the AI system would be built by people and, well you know the rest of it.

[mendis]

My original point was that even if we had an AI system checking for off track excursions, etc., as suggested earlier in the thread, it would still be liable to biases and errors just as the current human adjudicators are - because the AI system would be built by people and, well you know the rest of it.

Like I said, the only way to achieve consistent performing systems, is to evolve them. They won't be perfect for sure to begin with, but it will get there. Accepting human errors and emotions in governance of the sport would be farcical.

[MadMax]

My original point was that even if we had an AI system checking for off track excursions, etc., as suggested earlier in the thread, it would still be liable to biases and errors just as the current human adjudicators are - because the AI system would be built by people and, well you know the rest of it.

Like I said, the only way to achieve consistent performing systems, is to evolve them. They won't be perfect for sure to begin with, but it will get there. Accepting human errors and emotions in governance of the sport would be farcical.

Sport is a human undertaking. Emotion is inherent in sport - enjoyment is an emotion - and errors are part of sport - it's how the best beat the rest: by simply making fewer mistakes in many cases.

How long are you willing to take to get the system to be as good as a human? And when do you decide that they're ready to go "live"? How do you deal with the issues where the AI makes a call and people look at the same evidence and clearly see a mistake?

Why not just go the whole way and have the cars driven by AI and any transgressions likewise judged by AI? Remove the human element altogether. Its an obvious extension of your position, after all.

[mendis]

My original point was that even if we had an AI system checking for off track excursions, etc., as suggested earlier in the thread, it would still be liable to biases and errors just as the current human adjudicators are - because the AI system would be built by people and, well you know the rest of it.

Like I said, the only way to achieve consistent performing systems, is to evolve them. They won't be perfect for sure to begin with, but it will get there. Accepting human errors and emotions in governance of the sport would be farcical.

Sport is a human undertaking. Emotion is inherent in sport - enjoyment is an emotion - and errors are part of sport - it's how the best beat the rest: by simply making fewer mistakes in many cases.

How long are you willing to take to get the system to be as good as a human? And when do you decide that they're ready to go "live"? How do you deal with the issues where the AI makes a call and people look at the same evidence and clearly see a mistake?

Why not just go the whole way and have the cars driven by AI and any transgressions likewise judged by AI? Remove the human element altogether. Its an obvious extension of your position, after all.

I wouldn't complain if that happens. I am a misanthrope. The world needs release from human clutches.

[Zynerji]

Folks here speak as if these AI systems are not already mature and proven.

A high-res camera on corners to detect track limits violations is miles behind in difficulty compared to current state of the art.

[MadMax]

Folks here speak as if these AI systems are not already mature and proven.

A high-res camera on corners to detect track limits violations is miles behind in difficulty compared to current state of the art.

But they aren't infallible, which is the point being made.

[Zynerji]

Folks here speak as if these AI systems are not already mature and proven.

A high-res camera on corners to detect track limits violations is miles behind in difficulty compared to current state of the art.

But they aren't infallible, which is the point being made.

Every Amazon distribution center in America is 30% line following robots. That's millions of hours per year of infallible usage (failures of the systems are mechanical, not software).

We are talking a Python library and 50 Raspberry Pi4's with the HD camera and a 5G connection. The trick would be to put 3 cameras per corner from divergent angles, and let them come to consensus before reporting.With an F1 sized budget, they could do LiDar and even 3d map each corner in real time. This is enthusiast level stuff. Could be up and running in Mexico by Friday.

Going from politically whimsical to 99.999% accuracy for less than $2000USD is ultimately acceptable. Even if all it does is alert race control and provide pictures, it would help immensely.

[SiLo]

It's slightly baffling that they don't just use sensors within the plank area of the car and a strip of them at certain corners. Doesn't matter if all 4 wheels are off track depending on the angle of the car, as long as the tracking is consistent then there won't be any arguments for it.

[diffuser]

Folks here speak as if these AI systems are not already mature and proven.

A high-res camera on corners to detect track limits violations is miles behind in difficulty compared to current state of the art.

But they aren't infallible, which is the point being made.

Not sure why we're calling this AI but it's obvious to me that you can build a system to do a complex task and get it right every time. it may fail a few times at first but if you let the engineers improve it after every failure, it will approach perfection.

I think the biggest problem isn't making a system but agreeing on the rules. We see this over and over again that the local FIA teams interprets rules with more or less leaniancy. We see that the Budget CAP doesn't have a hard and fast penalty for going over the budget. There is just way too much grey area in the rules. When push comes to shove teams try to exploit that grey area and you end up in a controversy.

[MadMax]

It's slightly baffling that they don't just use sensors within the plank area of the car and a strip of them at certain corners. Doesn't matter if all 4 wheels are off track depending on the angle of the car, as long as the tracking is consistent then there won't be any arguments for it.

They already use a similar system for jump start detection so it's certainly not a new idea. But even jump start detection is opening to contentious situations - c.f. Bottas's most famous start.

The simplest answer is to just put something in the way - either a physical object or a road surface that no one wants to get their tyres on. A strip of wet astro turf is my favorite idea. You want to go outside track limits? Then be prepared to have a "moment". Or some nice foam/plastic marker. Hit a marker in qualifying and you lose your lap time. Do it in the race and you get a 5s penalty. And we can all move on.

[Zynerji]

It's slightly baffling that they don't just use sensors within the plank area of the car and a strip of them at certain corners. Doesn't matter if all 4 wheels are off track depending on the angle of the car, as long as the tracking is consistent then there won't be any arguments for it.

They already use a similar system for jump start detection so it's certainly not a new idea. But even jump start detection is opening to contentious situations - c.f. Bottas's most famous start.

The simplest answer is to just put something in the way - either a physical object or a road surface that no one wants to get their tyres on. A strip of wet astro turf is my favorite idea. You want to go outside track limits? Then be prepared to have a "moment". Or some nice foam/plastic marker. Hit a marker in qualifying and you lose your lap time. Do it in the race and you get a 5s penalty. And we can all move on.

You think digging up tracks and implementing a disposable bollard system is really more simple than commodity cameras and available software?

[MadMax]

It's slightly baffling that they don't just use sensors within the plank area of the car and a strip of them at certain corners. Doesn't matter if all 4 wheels are off track depending on the angle of the car, as long as the tracking is consistent then there won't be any arguments for it.

They already use a similar system for jump start detection so it's certainly not a new idea. But even jump start detection is opening to contentious situations - c.f. Bottas's most famous start.

The simplest answer is to just put something in the way - either a physical object or a road surface that no one wants to get their tyres on. A strip of wet astro turf is my favorite idea. You want to go outside track limits? Then be prepared to have a "moment". Or some nice foam/plastic marker. Hit a marker in qualifying and you lose your lap time. Do it in the race and you get a 5s penalty. And we can all move on.

You think digging up tracks and implementing a disposable bollard system is really more simple than commodity cameras and available software?

It removes any human element in interpretation or issuing of penalties - you drive on a slippery surface then you automatically lose time. The drivers will just avoid driving out there which is what is desired, isn't it?

And it doesn't have any tech to go wrong at an inopportune moment.

Cameras and software carrying out assessments is just several layers of complication that isn't required. And cameras can / do fail along with other issues that one can readily see can degrade the system.

Make any system as simple as possible.

[dans79]

Folks here speak as if these AI systems are not already mature and proven.

A high-res camera on corners to detect track limits violations is miles behind in difficulty compared to current state of the art.

But they aren't infallible, which is the point being made.

Every Amazon distribution center in America is 30% line following robots. That's millions of hours per year of infallible usage (failures of the systems are mechanical, not software).

There is a big difference between a well controlled warehouse space and an f1 track where a lot more unknowns can pop up.

for example.
https://www.caranddriver.com/news/a3070 ... -messages/