Artificial Information -the new How To Authority???

[BRENT in 10-uh-C]

I am sure many of you have heard about the phenomenon called Artificial Intelligence (AI) that is projected to soon take-over critical thinking, ...and possibly even common thinking.

So I have been wondering to myself for awhile about where will this information come from as it applies to Model-As. Maybe I am wrong, however it has been my view that mis-information about Model-As is 'beyond rampant' on social media platforms (-such as YouTube and FaceBook for starters) to the point now where I often shake my head in disbelief at some of the advice that is given. Ironically though, in several other special interest groups (Professional Engine Rebuilder group, Auto Shop Owners Group, RV Repair groups, Professional Auto Upholstery groups, Painting groups, et/al) I read the comments where the true (those that have vetted experience) professionals are complaining about the overwhelming amount misinformation that is being written, -or video content that is broadcast on these different platforms that is making life difficult for these professionals trying to educate & correct their customer about their preconceived ideas about something that was garnered because they read/heard/saw something on a social media platform that was false.

So, exactly where do we feel the information about Model-As will be found by AI in the future? Will this information be gathered from social media platforms -or somewhere else? Does the Restoration Guidelines & Judging Standards even have a chance in the AI world?? Thoughts???


.
.

[Pete]

[QUOTE=BRENT in 10-uh-C;2237646]I am sure many of you have heard about the phenomenon called Artificial Intelligence (AI) that is projected to soon take-over critical thinking, ...and possibly even common thinking.

So, exactly where do we feel the information about Model-As will be found by AI in the future? Will this information be gathered from social media platforms -or somewhere else? Does the Restoration Guidelines & Judging Standards even have a chance in the AI world?? Thoughts???./QUOTE]

The info will come from the same IQ level 50 that are propagating it now.
(such as)

"In the future all model A's will be required to have a 2300 Pinto engine in order to be on public highways."

[jb-ob]

Imagine 10 people driving the same type of vehicle that has the same navigation app. going to the same place. Their route would be the same.

No longer need for over priced universities. The same solutions will come from the same data from the same computers from scientists solving the same problem.

AI can not harness a small boy's ego, ie. bragging rites.

As long as there are uninformed listeners wishing to listen to uninformed story tellers, there will be whoppers.

What hobby has such a large audience of unskilled tinkerers that over rate their skills while disrespecting the geniuses that designed & built these cars in a world and for a world we never experienced.

My $.02, your mileage may & will vary. jb

[BRENT in 10-uh-C]

Originally Posted by jb-ob

What hobby has such a large audience of unskilled tinkerers that over rate their skills while disrespecting the geniuses that designed & built these cars in a world and for a world we never experienced.

Amen Jim!!

[Jim/GA]

Brent, I went into ChatGPT (a popular Generative AI text website) and asked a simple question:

"How do I adjust the valves on a 1929 Ford Model A engine?"

I got a horrible answer.

They have the option to regenerate the answer, which I chose.

The 2nd answer was worse than the 1st. I did it 3 more times. The answers were all useless.

The only good advice included was "It's always a good idea to consult a Model A engine manual or seek guidance from an experienced mechanic for accurate instructions tailored to your engine." The rest of it was garbage.

Every now and then I copy/paste a question I see here into ChatGPT to see what answer it comes up with. It has not been correct a single time.

[Hitman]

Sponsored Links (Register now to hide all advertisements)

I use AI in work, it's a great tool for what it is. It's always learning, so it continues to improve each day, and with each input. It will respond to your query appropriately. Ask it a direct question, and it can usually provide the correct response. Ask it a vague "what's the history of the Ford Model A in 2000 words" and it will pull in some errors or info you deem irrelevant.

I just asked ChatGPT "what's the correct wheel size for a 1931 Ford Model A", and it responded correctly. I followed up with "what carburetor does a 1931 Ford Model A have?" It responded: "The 1931 Ford Model A originally came with a Zenith 1-barrel updraft carburetor". It's not wrong.

Ask that of google and it will just provide you a list of what it thinks are relevant websites that can answer the question.

For fun, ask it to write a rap or country song, and it will provide one! Be careful though, it's always learning and never forgetting, so don't ask personal questions in AI.

[ronn]

as with all information- it will hit its bumps in the road.

the term garbage in and garbage out comes to mind.

everything is only as good as the accuracy of its input. Statistics is a perfect example. We can "bend" stats into any direction we wish. Its done and quoted daily. whether true or not.

[The Master Cylinder]

AI scares me, I keep thinking of the Terminator movies.

[jb-ob]

Again, who sourced the information in AI? What are their inherent biases ?

RE Model A, all information sourced from the Benson Achieves ?
Once randomly pulling blue prints I stumbled across a print for a wiring harness to use '31 Town Car Delivery side lamps on a '31 Deluxe Pickup.
Well at least Ford was prepared.

Beyond Ford most of the internal records from the Briggs Body Company appear lost (Detroit Public Library??). Briggs assemblied approximately 16,000 Cabriolet bodies in 1929. I'd like to believe I have substantial collection of copies of Ford America & Ford Canada literature regarding the '29 Cabriolet. No where is the popular '29 color combination of Rock Moss & Vagabond Green mentioned or offered. I've discovered 5 cabriolets that were originally painted this combo.

In both cases, whose information do we use?

Does AI now replace Moses delivering the Ten Commandants ?

[California Travieso]

Originally Posted by Jim/GA

Brent, I went into ChatGPT (a popular Generative AI text website) and asked a simple question:

"How do I adjust the valves on a 1929 Ford Model A engine?"

I got a horrible answer.

They have the option to regenerate the answer, which I chose.

The 2nd answer was worse than the 1st. I did it 3 more times. The answers were all useless.

The only good advice included was "It's always a good idea to consult a Model A engine manual or seek guidance from an experienced mechanic for accurate instructions tailored to your engine." The rest of it was garbage.

Every now and then I copy/paste a question I see here into ChatGPT to see what answer it comes up with. It has not been correct a single time.

Jim,

I’ve never used ChatGPT and I’m not inclined to do so in it’s initial phase. Like with a new car model, it’s best to wait until some of the bugs are worked. It might also be necessary for someone or some group with knowledge of Model A’s to give AI a head start or point it in the right direction, so to speak.

I think it would be necessary for someone to program an initial amount of information at the beginning in order for it to be useful for the Model A hobby, such as the books by respected mechanics, Judging Standards, the technical Bulletins, manuals such as the Andrews series, etc. it might also be a good idea to identify some people who it could rely on for good information.

AI learns quickly, I just hope it happens in my life time! At least it might be very useful for people who inherit our machines and stash’s of parts.

David Serrano

[Brentwood Bob]

Something to ponder.

[alexiskai]

I have a project in mind to train an AI on back issues of the Restorer, the JS, other known-good info sources. I don’t have access yet to the plug-ins I need to move forward on it.

[ursus]

EBay already provides a good example of how AI might be implemented to the detriment of our hobby. Any search for specific Model A parts may be rewarded with EBay listings for various items purporting to "fits Ford Model A" but could be for anything automotive, even as remote as for a 1988 Yugo. Typical is the complete engine kits that claim to be for a Model A but are actually for a late model 4-cylinder engine.

[Keith True]

I grew up playing around cows,so the term,AI means something completely different to me.If the neighbor was out working he would recruit me to point out to the AI lady which cows she was there to service.

[Oldbluoval]

KT….that’s about what’s happening with the “new AI”!!!!

[California Travieso]

Originally Posted by alexiskai

I have a project in mind to train an AI on back issues of the Restorer, the JS, other known-good info sources. I don’t have access yet to the plug-ins I need to move forward on it.

Alexiskai,

Is it a case of availability of plug-ins, cost or both? If it’s cost, maybe Barners could help.

David Serrano

[Hitman]

Originally Posted by alexiskai

I have a project in mind to train an AI on back issues of the Restorer, the JS, other known-good info sources. I don’t have access yet to the plug-ins I need to move forward on it.

The one you want is probably there, you need the ChatGPT 4 subscription to use them. Nearly 100 plugins were recently announced.

[alexiskai]

Originally Posted by Hitman

The one you want is probably there, you need the ChatGPT 4 subscription to use them. Nearly 100 plugins were recently announced.

I have the subscription, I need Code Interpreter, which is still in beta.

[DkFordor]

Originally Posted by ronn

the term garbage in and garbage out comes to mind.

This is particularly true!

AI, or machine learning, needs to be trained on a dataset prior to use. Think of the early spam filters, where certain terms would trigger a 'spam'-label. Spammers would then eg write 'f*ck' to avoid this. The new generation of spamfilters are AI-based. So when you tell your mailbox that this mail was in fact spam - you in fact train the AI behind it.

This also potentially leads to 'dataset' bias. Upon request early AI algorithms would create paintings of persons; ask for a man you'd typically get a man in suit and tie - ask for a woman you'd typically get swimsuits, bikinis or the like

Trained on a proper dataset I am sure AI could prove a valuable asset for the hobby - particularly for the non-experts!

[updraught]

Last time I mentioned having the judging standards in digital form I got hammered.