I have finally gotten around to getting the aftermarket carpet into my 2002 CETA. The front half is now trimmed and sitting in there right with the only part being putting things back together. Well I found out some of the containers I was using to hold the screws had been tipped over onto the back of the workbench and all the screws are not mixed up (console, dash, door trim, sail panel, seat belt, etc.) While the trim panels are mostly universal (some are far longer than others) there are some with huge washers, small cone shaped washers, and tapered heads, flat heads, hex heads, etc. I can't for the life of me remember from 3 years back what goes where anymore and of course I didn't take pictures.

So I asked AI what screws hold down the center console and got this back as a reply. Pretty nifty is accurate but I want to check and confirm as AI at work is about 25% accurate.

 

I had GROK do some camshaft comparisons, Methanol flow rates and horspower vs boost estimates on gen 3 5.3L
GROK really liked the combination:
Big Truck Torkinator .600" lift 212 intake/ 218 Exhaust duration with VSR Billet 78/75 pushing 190kPa and 10 gph of 70% methanol.
..Estimated 580 wheel horsepower through 4L80 and AUTO transfer case.

 

If AI is picking up what people post in general, hopefully it goes back about 10+ years ago before people jumped in with what they "think" how **** works instead of how it actually works.

 

""hopefully it goes back about 10+ years ago""

It appears that " It Don't".
seems like the most current bullshit is prioritized, and the older bullshit is the less it is referenced.
In our hobby, this means all the **** we DID, wrote, learned- or searched for to do our Gen 3 and Gen 4 mods 10 to 15 years ago is on the verge of being lost to the sand of time.
Inverse Square law of Historical importance is applied:

 

I don't know how much I trust AI. It just seems fake to me.

 

It is fake, hence the name. Artificial "Intelligence".

I quoted the last part for sarcasm, I'm sure I'm scheduled for execution now.

 

We (humanity in general) aren't ready for "artificial intelligence" until we get rid of all the "natural stupidity".

 

I think people's expectations on what AI can do is way too high. The largest issue I am seeing with it from my work is the limits of what is used to "train" it. The limits of its ability to filter and ignore input make it really inefficient in training. You have to be super specific is what you use else it makes crap up or "hallucinates" and give bogus output. The results I got were based off of commerce sites and this site mostly from the references Duck Duck Go provided. That is nice as it gives me some insight into how it came up with the answer. At this stage it is a step or two above a standard web search but that is about it. I'd trust it in giving me tire sizes, spark plug recommendations, and torque specs, maybe event fuel trim recommendations for specific outcomes but I wouldn't trust it for much more beyond that. I think for car stuff, gun stuff, camera/video stuff it will be great in going through the mountains of BS on the web, but as earlier stated, if it weighs more recent content above the older just because its newer it won't be authoritative anytime soon.

 

and on August 29th, Skynet became aware

 

This is the greatest challenge for AI to overcome, and I really don't see how (in the foreseeable future) it will be able to do so. For things like factual stats (production numbers, factory service information, originality or factory content/configuration, historical data on older modifications or service replacements, etc.), the information rarely improves over time; usually, in this case, the best data is the oldest data because that is when it was fresh and also subject to the greatest amount of accurate peer-review. But then for other types of data, such as the latest standard in camshaft design, valve spring upgrades, cylinder head & intake options, etc. - all of this is likely going to be best answered with the latest data as these things continue to evolve and improve (would you really want AI to give you a recommendation on "best LS valve springs" primarily using data from 2002? LOL!) So how do you train AI to know when the older data is more accurate/relevant vs. not? If they can work that out, then it's useful. Otherwise it's really only a tool for information or topics that truly evolve and/or improve over time - beyond that, you'd be better off digging for the old data.

 

There is No such thing as AI. It’s a fancy compiler that has existed since the beginning of computers in one form or another.

‘Nothing new happening, but a whole bunch of people have been fooled into believing it.

 

There are many different definitions of "intelligence", but one could certainly argue that various types of computers (or software) have been a form of AI for many decades now. I agree that AI is nothing new, it's just a newer term for the latest advances in computing.

 

A.I. stands for Artificial Incompetence

Here is a good example. When Alpha Go was "trained" to play the game Go, it got to where it could defeat a Grand Master human player. Then an engineer started poking at the AI program and found a strategy that won 19 times in 20, a strategy that you or I could defeat when we had never seen a game of Go before and only had 30 seconds of explanation as to what the objective of the game was.

In truth, an AI doesn't actually understand what it is doing. It's a complex If/Than logic matrix.