its amazing how often were being posts here basically saying

" I tried this AI engine and it gave me inconsistent results "

two thoughts
a, I feel link just saying , are you asking us to debug the AI or whats your question.
b. im I two minds , at least people are checking two AI engines , but what happens when the AI learn off each other ?

any thoughts ?

 

Unfortunately, this is one of the consequences of AI slop.
Writing which is inaccurate, misleading, factually incorrect, with minimal review, which will eventually lead to an internet full of junk content.

My worldview has become a pair of notches more depressing because of this.

 

I hav no confidence in AI because there are no references to where its information is coming from. Copilot is particularly intrusive and offers very incorrect answers. I would turn it off if I could find a simple way to do it.

 

its amazing how often were being posts here basically saying

" I tried this AI engine and it gave me inconsistent results "

two thoughts
a, I feel link just saying , are you asking us to debug the AI or whats your question.
b. im I two minds , at least people are checking two AI engines , but what happens when the AI learn off each other ?

any thoughts ?

When they learn off of each other the result is even worse AI slop.

https://en.wikipedia.org/wiki/Model_collapse

Impact on large language models
In the context of large language models, research found that training LLMs on predecessor-generated text — language models are trained on the synthetic data produced by previous models — causes a consistent decrease in the lexical, syntactic, and semantic diversity of the model outputs through successive iterations, notably remarkable for tasks demanding high levels of creativity.[25]

 

I hav no confidence in AI because there are no references to where its information is coming from. Copilot is particularly intrusive and offers very incorrect answers. I would turn it off if I could find a simple way to do it.

ChatGPT will provide both references and a narrative of rationale behind the output.

ChatGPT can be exceedingly helpful but it is like active lane maintenance and using it as a self driving facility will end in tears. Effective use requires constant interactive refinement and enough knowledge in the user's hands to vet the results.

 

ChatGPT will provide both references and a narrative of rationale behind the output.

ChatGPT can be exceedingly helpful but it is like active lane maintenance and using it as a self driving facility will end in tears. Effective use requires constant interactive refinement and enough knowledge in the user's hands to vet the results.

And like active lane guidance , people are taking AI as "true" , and quoting it !
Wonder what's going to happen
I see options of
We humans learn to use AI within it's limits
AI gets clever enough to cope with humans
Ai is banned
Ai takes over the world ..

 

And like active lane guidance , people are taking AI as "true" , and quoting it !
Wonder what's going to happen
I see options of
We humans learn to use AI within it's limits
AI gets clever enough to cope with humans
Ai is banned
Ai takes over the world ..

You left out: AI, like telecommunications fundamentally changes the world and fades into the background as "just another tool" that we integrate into our lives. The ethical distortions and lost skills become invisible, having faded into obscurity, and the world keeps spinning, with humans on it, cycling between hope and despair, between progress and regress until something bigger than us wipes us out.

It's not about learning limits, it's about living in the context that the world as a whole presents. Humans can be very clever, and very stupid—in aggregate we race to the bottom because the top is complicated and expensive.

 

Effective use requires constant interactive refinement and enough knowledge in the user's hands to vet the results.

And therein lies a major drawback. Unfortunately, the very people who perhaps could benefit most from AI and would rely on it, do not have that knowledge. Using AI to gain knowledge about a subject means you already need to know enough about that subject to tell when the AI is telling lies. Chicken and egg.

 

back to the original question,
on



"
its amazing how often were being posts here basically saying

" I tried this AI engine and it gave me inconsistent results "

two thoughts
a, I feel link just saying , are you asking us to debug the AI or whats your question.
b. im I two minds , at least people are checking two AI engines , but what happens when the AI learn off each other ?

any thoughts ?
"

 

ChatGPT can be exceedingly helpful but it is like active lane maintenance and using it as a self driving facility will end in tears.

Active lane maintenance is so dreadful that using it for active lane maintenance usually ends in tears!

 

Active lane maintenance is so dreadful that using it for active lane maintenance usually ends in tears!

Depends on the vehicle. Some newer versions are quite good.

 

That is correct.
The difference between my wife’s 2021 Honda and my 2025 Toyota is very significant.
Likewise with the radar-assisted cruise control.
However, unless it is a long highway stretch, I usually keep them off.

 

cars , are getting very clever

but im wondering how people using AI are going to affect the forums,
are we going to get yet more " this is what two ai disagree on say " type questions ,
will the forums become a chat place for old foggiest like me to reminisce
et cetera ..

 

I had a conversation with Claude.ai, the result of which was to correct Claude's "received wisdom" as its (his?) only knowledge was what could be found on the internet. I asked if Claude would incorporate the new fact into its knowledge base, and was told it was not possible. I asked if the new knowledge could be used if I posted it on AAC, and was told that would then be on the internet and could be used.
. . .I wish I could remember what the conversation was about. Needless to say, I never got around to posting in on AAC.

 

OpenAI now possibly the world's most valuable startup | AP News

I couldn't find the other article I saw this morning touting OpenAI as the largest capitalized company in the world. They will have to bring new atomic energy plants online just to power the monsters that host it. It now consumes more energy than a 100,000-population city! Horribly expensive just to turn the platform on before it even does anything.

 

These AI tools are valuable for people who aren't domain experts, I solved a painful problem yesterday that would have taken me a week to do using normal documentation searches and chat forum skimming. By interacting back and forth over the course of a an hour I got the solution.

My problem was specialized, it is to stamp metadata on a generated binary file, generated by the Visual Studio build system. In that metadata I want to record the git branch and commit ID at the time of the build. We use this to trouble shoot and so on, because if we get issues reported I can get that metadata from the .EXE and go straight to the exact code used to build that EXE.

This is esoteric stuff, requiring intimate knowledge of Git and MSBuild, I could have solved it myself but I know it would have taken days of effort.

Here's the first few interactions FYI








and so on, for about an hour or so...



eventually...



This is without doubt a very helpful system, saved me days of work and needless studying of docs that I will never read again. Faced with these kinds of challenges only a luddite would refuse to use AI this way.

 

I'm still a bit skeptical of what they now call AI. I spent my career in industry automating process controls from standalone flow, temperature, level, pressure, etc. and paper recorders/clocks. The controls I specified and installed were computer based and integrated. PID control was even further automated with PID Auto Tuners. Was that AI? The recording data could be manipulated and integrated into other applications. Was that AI? My son is doing the same thing in the business arena. Automating business transactions and processes. He tells me that they are using AI. How I have no idea. I hear of code monkeys using AI to speed up their writing of code and debugging it. Once again, I have no idea how but it is what they claim. AI is (or should be) more than just being able to parse a computer's response into a complete and grammatically correct sentence or finding the solution to a complex equation.

 

I'm still a bit skeptical of what they now call AI. I spent my career in industry automating process controls from standalone flow, temperature, level, pressure, etc. and paper recorders/clocks. The controls I specified and installed were computer based and integrated. PID control was even further automated with PID Auto Tuners. Was that AI? The recording data could be manipulated and integrated into other applications. Was that AI? My son is doing the same thing in the business arena. Automating business transactions and processes. He tells me that they are using AI. How I have no idea. I hear of code monkeys using AI to speed up their writing of code and debugging it. Once again, I have no idea how but it is what they claim. AI is (or should be) more than just being able to parse a computer's response into a complete and grammatically correct sentence or finding the solution to a complex equation.

I agree with what you say. AI is just software.
On the lighter side: There are so many gadgets in modern cars. Yesterday, I put my car in reverse and it played a video of someone getting run over!

 

Yep, I read the other day about a guy in an EV that got a ticket for loud mufflers. No kidding, the car actually had a noisemaker installed to mimic exhaust sounds.

 

My question was aimed at how AI was going to affect this web site ,