The Limits of AI

perplexity.ai ^ | 08/24/2025 | self

[Bob434]

i have a photographic memory, it’s just not developed yet

[PeterPrinciple]

AI reminds me of the kid at the hardware store who always has an answer but doesn’t know what he is talking about.

[verum ago]

This post and its contradictory 'I consulted my pet-AI' comments have turned into a magnificent example of why people shouldn't blindly trust 'AI' for answers.

Simply put, what the vast majority of people call 'AI' these days are LLMs- Large Language Models. It's glorified text prediction software, and nothing more.

There is no thought or evaluation process behind it. No self evaluation. No ability to recognize right or wrong. No objective ability to evaluate for correct or incorrect.

A quick Google Lens (non-AI) search gives correct, easily verifiable (and verification is an important step with these tools!) results.

Simply put, if enough people asserted online that 2+2=5, so would 'AI'. It has no actual 'intelligence' whatsoever, and I'm beginning to be embarrassed by how hard people lean on it on this website at times.

Just like the other day, when multiple posters on a thread about redistricting used AI to insist that every single US state has at least 1 Democrat in their House delegation, which 'fact' doesn't stand up to the sniff test, much less actually glancing at a roster of Representatives. But hey, they saved a few seconds by turning their brains off and letting corporations that they generally claim not to trust as far as they can throw to step in and do their own 'thinking' for them...

Suspending your own critical thought in favor of something literally incapable of such is not a good idea.

AI in its current state is a timesaver only for those who know how to, and will actually take a moment to vet the results. Everyone else will inevitably be lead astray by its equally inevitable errors.

[antidemoncrat]

Isn’t it kind of ironic that in the “Matrix” movies the
machine couldn’t get enough power so it started raising
human fuel to supply it? Now it’s likely the US probably
can’t produce enough power to satisfy AI’s needs. I wonder
if the Chinese recognized the importance of AI years ago
when they started building all those coal fired power plants
which they have a mucho supply of coal to run.

[cymbeline]

“We are still in the early days of AI.”

A huge amount of money is being spent on AI. What are the vast amounts of revenue being generated to pay the bills.

[Retrofitted]

It is. Here’s the full photo: https://www.pinterest.com/pin/618189486386512419/

I used AI to track it down by the microphone. And, man, is Lola gorgeous.

[aspasia]

Facial recognition seems to work to catch perps at customs, most of the time?

[oldtech]

Makes me wonder just why I would need AI as I can make mistakes just as easily on my own.

[Bob434]

Lol, same here

[13foxtrot]

Whatever. If the AI was trained/tuned to recognize old movie stars it would work without fail. Guess what, I didn’t recognize the actress either. I knew it wasn’t M Monroe, but if I was forced to guess, as was the AI, who knows what name I would have came up with.

AI is amazing tech, it has limitations and it has dangers.

I have been using it more and more with great success and satisfaction.

The danger is how it will effect those (younger folks) who grow up using it versus oldsters like me who developed expertise in various professions in the pre-AI world.

[Dr. Sivana]

but if I was forced to guess, as was the AI, who knows what name I would have came up with.

The AI was NOT forced to guess. It presented a bad guess as a fact. It would be better if it would present doubtful answers as such more frequently.

[13foxtrot]

That is the lesson. You can’t trust AI anymore than you can trust a human. I use AI a lot, it works fine. Trust but verify.

The same Morans that believe mainstream media will believe AI without question. It is the same as its always been.

[Dr. Sivana]

Yes. Agreed on all points.

[BigFreakinToad]

I somehow that will ever happen. I thought the current AI’s are passing the Turing test.

[cgbg]

The average human IQ is 100.

AI can meet that standard.

Lol.

[Dr. Sivana]

I thought the current AI’s are passing the Turing test.

It does in some areas. Not consistently. I should also add the qualifier that it should emulate an intelligent, very-well educated, honest, cooperative human being with exceptional skills in searching the Internet and access to same.