Is using design aids like AI and Machine Learning legal?

[Just_a_fan]

You can approximate physic also with AI/ML. With enough data, you can train a system that maps given 3D to CFD output.

But the highlighted bit is the important bit, isn't it? It's getting a large enough dataset that's the issue for a team, because their own data, whilst plentiful, will be inherently limited by their own ability to generate meaningful output.

I doubt any team has a large enough dataset for ML to generate wing profiles in lieu of CFD.

I would say every team has enough data probably. In 3D, even one pair instance contains very valuable information, if you can define a set of good regularizations, you can reduce the space of possible solutions, hence the requirement for more data becomes less of an issue.

I would make an arrogant statement here, with 4-5 good AI engineers, and good structured historical data of an average F1 team, you can build a system that generates designs that beat most of the current designers in one year.

A whole car or just a single component? Because I doubt the latter and I seriously doubt the former.

[Hoffman900]

You can approximate physic also with AI/ML. With enough data, you can train a system that maps given 3D to CFD output.

But the highlighted bit is the important bit, isn't it? It's getting a large enough dataset that's the issue for a team, because their own data, whilst plentiful, will be inherently limited by their own ability to generate meaningful output.

I doubt any team has a large enough dataset for ML to generate wing profiles in lieu of CFD.

I would say every team has enough data probably. In 3D, even one pair instance contains very valuable information, if you can define a set of good regularizations, you can reduce the space of possible solutions, hence the requirement for more data becomes less of an issue.

I would make an arrogant statement here, with 4-5 good AI engineers, and good structured historical data of an average F1 team, you can build a system that generates designs that beat most of the current designers in one year.

Yeah, not even close.

If the teams, and many other design/ engineering / product industries could could do that and reduce labor costs, they absolutely would. Maybe a few decades down the road, but this is just hogwash at this point.

Despite all the team’s data, and calibration, they still have correlation issues in CFD and FEA, it all needs a lot of human intervention still.

In my world, everyone talking adnaseum about AI were talking blockchain and “BIG DATA” a fe years ago, they just keep moving onto whatever the latest tech buzzword is.

[Zynerji]

But the highlighted bit is the important bit, isn't it? It's getting a large enough dataset that's the issue for a team, because their own data, whilst plentiful, will be inherently limited by their own ability to generate meaningful output.

I doubt any team has a large enough dataset for ML to generate wing profiles in lieu of CFD.

I would say every team has enough data probably. In 3D, even one pair instance contains very valuable information, if you can define a set of good regularizations, you can reduce the space of possible solutions, hence the requirement for more data becomes less of an issue.

I would make an arrogant statement here, with 4-5 good AI engineers, and good structured historical data of an average F1 team, you can build a system that generates designs that beat most of the current designers in one year.

Yeah, not even close.

If the teams, and many other design/ engineering / product industries could could do that and reduce labor costs, they absolutely would. Maybe a few decades down the road, but this is just hogwash at this point.

Despite all the team’s data, and calibration, they still have correlation issues in CFD and FEA, it all needs a lot of human intervention still.

In my world, everyone talking adnaseum about AI were talking blockchain and “BIG DATA” a fe years ago, they just keep moving onto whatever the latest tech buzzword is.

I've now removed 9 people from an office now that AutoGPT does their work flawlessly. I think going zero-human will be a big push in the future.

[dans79]

Scary question is… how long will it be before AI is able to do that sort of thing and return a good result.

As someone who has actually worked with neural networks and other AI like tech, it will be decades, if ever.

All the "AI" that currently exists is just a lot of conditional logic set up by the developers. No one has gotten anywhere close to true AI. True AI requires the system/application to learn why something is better.

AI today is just a glorified solver wrapped in a bunch of fluff.

[Just_a_fan]

Scary question is… how long will it be before AI is able to do that sort of thing and return a good result.

As someone who has actually worked with neural network and other AI like tech, it will be decades, if ever.

All the "AI" that currently exists is just a lot of conditional logic set up by the developers. No one has gotten anywhere close to true AI. True AI requires the system/application to learn why something is better.

AI today is just a glorified solver wrapped in a bunch of fluff.

That's the thing - it's all interpolating within existing datasets. To make advancements requires extrapolation and that's a whole different ball game.

[Hoffman900]

I would say every team has enough data probably. In 3D, even one pair instance contains very valuable information, if you can define a set of good regularizations, you can reduce the space of possible solutions, hence the requirement for more data becomes less of an issue.

I would make an arrogant statement here, with 4-5 good AI engineers, and good structured historical data of an average F1 team, you can build a system that generates designs that beat most of the current designers in one year.

Yeah, not even close.

If the teams, and many other design/ engineering / product industries could could do that and reduce labor costs, they absolutely would. Maybe a few decades down the road, but this is just hogwash at this point.

Despite all the team’s data, and calibration, they still have correlation issues in CFD and FEA, it all needs a lot of human intervention still.

In my world, everyone talking adnaseum about AI were talking blockchain and “BIG DATA” a fe years ago, they just keep moving onto whatever the latest tech buzzword is.

I've now removed 9 people from an office now that AutoGPT does their work flawlessly. I think going zero-human will be a big push in the future.

I don’t know what you do for work, but it’s not going to happen in engineering for physical products for a long time, at least stressed.

I mean, you’ve even admitted your replies here are based on conversations with ChatGPT and explains why most of the tech you share from that is just nonsense.

You can “AI” and 3D print some useless desk trinket.

At this point, AI is just for a bunch of MBA’s to trick other people into “investing” (aka: take their money).

[maygun]

But the highlighted bit is the important bit, isn't it? It's getting a large enough dataset that's the issue for a team, because their own data, whilst plentiful, will be inherently limited by their own ability to generate meaningful output.

I doubt any team has a large enough dataset for ML to generate wing profiles in lieu of CFD.

I would say every team has enough data probably. In 3D, even one pair instance contains very valuable information, if you can define a set of good regularizations, you can reduce the space of possible solutions, hence the requirement for more data becomes less of an issue.

I would make an arrogant statement here, with 4-5 good AI engineers, and good structured historical data of an average F1 team, you can build a system that generates designs that beat most of the current designers in one year.

A whole car or just a single component? Because I doubt the latter and I seriously doubt the former.

Single component, the whole car thing is a very different problem and I am sure we are not there yet or even close to it.
Ofc there would be specialists in the loop when using/designing this system even for the single component.

[maygun]

Yeah, not even close.

If the teams, and many other design/ engineering / product industries could could do that and reduce labor costs, they absolutely would. Maybe a few decades down the road, but this is just hogwash at this point.

Despite all the team’s data, and calibration, they still have correlation issues in CFD and FEA, it all needs a lot of human intervention still.

In my world, everyone talking adnaseum about AI were talking blockchain and “BIG DATA” a fe years ago, they just keep moving onto whatever the latest tech buzzword is.

I've now removed 9 people from an office now that AutoGPT does their work flawlessly. I think going zero-human will be a big push in the future.

I don’t know what you do for work, but it’s not going to happen in engineering for physical products for a long time, at least stressed.

I mean, you’ve even admitted your replies here are based on conversations with ChatGPT and explains why most of the tech you share from that is just nonsense.

You can “AI” and 3D print some useless desk trinket.

At this point, AI is just for a bunch of MBA’s to trick other people into “investing” (aka: take their money).

So you think all of this is bullshit and not useless: https://www.autodesk.co.uk/products/fus ... -extension

And this is a just general-purpose use case.

Ofc there is hype and some factually wrong and overselling ideas, but at this point AI/ML/Deep learning whatever you call, it can enhance/help engineering for physical products even today.

[Hoffman900]

I do because people take stuff like FEA all the time and cut material based on the pretty colors, and consequently said part fails. The AI is doing the same thing, it’s all “trained” on the same data. Autodesk is just using machine learning to do things, and the algorithms is trained on data that may or may not be good. Considering all the issues CFD and FEA have correlating to the real world, the datasets are dubious at best. Labeling AI is just marketing as selling licenses is how they make their money.

CFD is the same. We’re in 2023 and teams as sophisticated as Mercedes, with all their computing, expertise, wind tunnel, and real life measurements still have correlation issues. “AI” is going to have the same.

+1 to what Dans79 said. AI today is not artificial or intelligent and whatever is commercially available today isn’t what you’re making it out to be.

[dans79]

So you think all of this is bullshit and not useless: https://www.autodesk.co.uk/products/fus ... -extension

That is basically a glorified solver. minimize or maximize variable A by systematically varying variables B, C, D etc. it's not efficient at all when the problem gets complex.

[Hoffman900]

So you think all of this is bullshit and not useless: https://www.autodesk.co.uk/products/fus ... -extension

That is basically a glorified solver. minimize or maximize variable A by systematically varying variables B, C, D etc. it's not efficient at all when the problem gets complex.

This.

And said part it designs still may not validate to the real world.

[maygun]

I do because people take stuff like FEA all the time and cut material based on the pretty colors, and consequently said part fails. The AI is doing the same thing, it’s all “trained” on the same data.

CFD is the same. We’re in 2023 and teams as sophisticated as Mercedes, with all their computing, expertise, wind tunnel, and real life measurements still have correlation issues.

The wind tunnel is a restricted game, due to restrictions you cannot get a good approximation. CFD, simulation etc also another approximation game.

I dont agree with the 'trained' on the same data thing. Yes, the models have biases regarding their training data, but with enough data and proper settings, models can generate out-of-distribution examples as well. I am also mostly on the same side with you on the debates regarding the capabilities of current ai/ml models, but in this case, I think you are underestimating the current capability of the state-of-the-art.

When it comes to state-of-the-art software design, simulation, machine learning, and data science, I am sure that F1 teams are nowhere close to it. Right now they dont care, and I understand why they dont care (with the budget cap you cannot invest in something that you are not sure if it is going to work or not), but as soon as one team do a proper investment, all of them going to follow it.

[maygun]

So you think all of this is bullshit and not useless: https://www.autodesk.co.uk/products/fus ... -extension

That is basically a glorified solver. minimize or maximize variable A by systematically varying variables B, C, D etc. it's not efficient at all when the problem gets complex.

On the product page, they say they use ML/Generative AI to generate alternative design, do you think they are lying?
Btw the useful ml product that is in my head for F1 design is also a 'glorified solver'.

Your designer creates the initial design. Set the variables that need to be optimized (A), but the cardinality of varying variables is huge, hence impossible to vary them one-by-one, and the new AI/ML program does these steps way efficiently than current glorified solvers.

[dans79]

On the product page, they say they use ML/Generative AI to generate alternative design, do you think they are lying?
Btw the useful ml product that is in my head for F1 design is also a 'glorified solver'.

I've used Autodesk products for over two decades, good/strong/correct marketing has never been their strong suit. My degrees is in computational physics, and I work as a systems engineer. Their is a massive difference between what I call a system I've created and what the sales and marketing department call it.

Your designer creates the initial design. Set the variables that need to be optimized (A), but the cardinality of varying variables is huge, hence impossible to vary them one-by-one, and the new AI/ML program does these steps way efficiently than current glorified solvers.

The issue is the AI is "dumb" it will go down every rabbit hole following gains. it doesn't learn from one run to the next, what is conceptually good, and what is conceptually bad like a human does. By that I mean 6 months down the road when a similar project comes up, it doesn't carry forward any of conceptual knowledge it learned on previous projects.

This is one of the reasons why "AI" is a heavily abused term, just like "cloud" was a decade ago.

[maygun]

On the product page, they say they use ML/Generative AI to generate alternative design, do you think they are lying?
Btw the useful ml product that is in my head for F1 design is also a 'glorified solver'.

I've used Autodesk products for over two decades, good/strong/correct marketing has never been their strong suit. My degrees is in computational physics, and I work as a systems engineer. Their is a massive difference between what I call a system I've created and what the sales and marketing department call it.

Your designer creates the initial design. Set the variables that need to be optimized (A), but the cardinality of varying variables is huge, hence impossible to vary them one-by-one, and the new AI/ML program does these steps way efficiently than current glorified solvers.

The issue is the AI is "dumb" it will go down every rabbit hole following gains. it doesn't learn from one run to the next, what is conceptually good, and what is conceptually bad like a human does. By that I mean 6 months down the road when a similar project comes up, it doesn't carry forward any of conceptual knowledge it learned on previous projects.

This is one of the reasons why "AI" is a heavily abused term, just like "cloud" was a decade ago.

Thanks for the insights about the real design part. Yeah, i agree cloud/ai is used as a buzz/hype word.

But I dont agree with the dumb part. Transfer learning is a proven concept in various domains, and I don't see why it wouldn't help here also. Ofc these systems cannot learn 'real' concepts and transfer to other problems as humans do, but they also dont start from scratch.

Today for even dumb hyperparameter search problems you can build ml methods that beat random/bayesian search algorithms. I am not claiming we can have a program that we would give 'design a from wing' and it would give us a pareto-efficient design. but it would accelerate or even gonna help the designer to be more creative when they design an element.