Let’s use AI for tuning.

Anyone else been thinking or trying to use tune and log data pushed through AI to build maps yet? I’m sure it’s the way of the future.

The last two, or three, generations of consumer gaming/desktop GPUs support a degree of "Deep Learning", the problem would be programming the software to utilise it. Heck, for the rather simple application, I wouldn't be surprised if modern "smart" 'phones had enough processing power for the job.

Even then, the limited number of variables may be best contolled by traditional tables and corrections.

I think it's already been done for some time at industry level. For a privateer/small operator to develop their own package if nor wanting to pony up for AVL or similar you would need to be doing many and varied engine package tunes to return the time invested, the reality is after you have done one of any particular type you are going to have a base to work from anyway.

I think there may be a couple of free or very cheap basic engine simulations where you can input, syroke/bore compression ratio, valve arrangement etc and get a predicted timing valve.

https://www.avl.com/en/simulation-solutions

I believe mainline dynos will communicate with some aftermarket ecus, you could probably get Ioannis from Plex to write feedback for ignition and fuel trims if it isn't already possible from the PCA-2000.

Thanks for your opinions and information. This is sure to be a hot topic as the tech develops.

I agree it's an interesting topic. My experience has been most cars have issues needing diagnosed which would send AI off into the weeds, but I could see it as another valid tool in a skilled calibrator's toolbox. As long as human beings are still building cars and making parts, we'll make enough mistakes that a human being will need to sort out what's going on. I believe that will give me a long glide path of job security.