[gassman_67]

I was trying something, I used AI to assist insetting up my 461 pro flo 4 tune. I put in car, trans,stall, diff ratio,distributor type, heads CR,and cam part number. Interesting, AI gave me a lot of info. What the heck, I saved my current tune, and tried it. Some of the data was very close to what I had been using, some different. I plugged it in and tried it. Very interesting results

[62posbonny]

All AI does is filter through the internet for information pertaining to your query and regurgitates it back to you in a nice tidy package. Someone, somewhere, likely on a forum much like this one has taken the time to post their results on an engine similar enough to yours and the AI clanker gave you their knowledge as its own. It's likely that the info you received will get you close, so i'm not surprised. I prefer to go straight to the source, though. By the way, these Pro Flo 4 setups really do run good for how simplistic their tuning maps are!

[gassman_67]

true, they do run good. each engine combo is slightly different tho, just kinda neat , that's all

[dataway]

I basically use it as a search tool for hard to find items or information. I searched forever for the allowable axial clearance for a Kubota KX040 excavator boom pivot pin, no where to be found, Chat GPT found it.

It's really good at correlating tons of info into something usable. Can save you many hours of searching.

[JLMounce]

I've been using it a bit, more specifically with my transmission tuning. Because it scrapes everything from forum posts, to OE literature and design schematics for basically every engine in existence, it can get pretty close to what you need.

Nothing is going to beat putting the car on rollers and doing a cell by cell MBT tune along with a road tune for driveability. But for the types of stuff most of us are working with here, AI is going to get you pretty close.

[68azbird]

Originally Posted by JLMounce

I've been using it a bit, more specifically with my transmission tuning. Because it scrapes everything from forum posts, to OE literature and design schematics for basically every engine in existence, it can get pretty close to what you need.

Problem with that is that if the data is wrong to start with the end result wont be as accurate as you think.
It is the same computer problem from the beginning. Garbage in, Garbage out.

[JLMounce]

Originally Posted by 68azbird

Problem with that is that if the data is wrong to start with the end result wont be as accurate as you think.
It is the same computer problem from the beginning. Garbage in, Garbage out.

That can happen with or without AI. I'm not defending AI here, but I can literally post anything I want on this forum and somebody could be mislead by the information. It's the same situation.

Most of the LLM's have access to documentation that many of us simply don't. OE specs and technical data, books and manuals that are out of print or we simply don't know about.

You have to understand your own ability to think through a problem and prompt the machine to give you the information you're looking for. And, when it hallucinates, you have to have the discernment to understand that the information you're receiving doesn't make any sense.

Some of the models are inherently agreeable, even to the point of not being factual. You have to make sure you're probing for information in a non-biased way, because the model can and does find the information necessary (accurate or not) to agree with you.

Point being, is that you have to take the same precautions with AI as you do if you're simply coming to a forum, or asking questions on a FB group.