[PeterPrinciple]

Most other tasks, including naming, attention, language, and abstraction were performed well by all chatbots.

---

think about that.......................

problem solving, not so much.

[Twotone]

If AI learns from us, & we are aging & declining mentally, it stands to reason that the same would happen to AI. ;-)

[HombreSecreto]

Seems like there’s a difference between learning and discernment... well, duh.

[wildcard_redneck]

They need to develop AI that can forget useless information.

[Deaf and Discerning]

I use Microsoft’s Copilot often, but only as a super power search engine, to provide me with resources that MIGHT support Copilot’s responses to my questions. Some of the website links that Copilot suggests are totally irrelevant to my questions, but it does a goood job of chcking my spelllling and also my grammmer too...

[libertylover]

I don’t think I have EVER connected to a customer service chat line to help me with something and have the chatbot actually help, but normally you have to go through the chatbot to get to a human who can help you.

[imardmd1]

If they “knew” as much of the mass-time-dimension facts as did Albert, could they ever deduce the concept that light shows the special inescapable relationship between mass and power?

[coloradan]

“The study also found that older versions of these chatbots, much like aging human patients, performed worse on the tests.”

This is what passes for journalism now. The quote implies that any given chatbot loses ability with time (”much like aging human patients”) instead of simply stating that newer ones are more advanced and score higher.

[I-ambush]

Maybe the problem is knowledge without intellect, a which has always been attributed solely to God, angels and humans. I never thought that inanimate human creations could truly reason, no matter how many processors or how fast the clock speed.
Interestingly, I watched a documentary on YouTube last week comparing NASA and SpaceX. I figured it must have been AI generated when it referred to a NASA official receiving an email in 1966.

[Harmless Teddy Bear]

Almost all leading large language models, or “chatbots,”

Well. Isn't that interesting.

Once their new god is failing they admit what I have been saying all along.

It is not AI.

I am shocked.

[KarlInOhio]

Are the older versions performing worse than they did last year? That would indicate cognitive decline. Or do the older models performing worse than the newer models? That would indicate cognitive improvement, like comparing results from a five year old versus a six year old. The six year olds do better, but that is not evidence of the five year olds suffering from a decline.

Maybe the neurologists giving the tests are the ones suffering from a decline by not discussing that.

[LittleBillyInfidel]

If an LLM is initially trained on high quality data sets and input - but later programmed to absorb any/all open source data available on the internet (like, say, a Saturday night blog post by “NextRush”), then it’s bound to show severe cognitive decline.

[mairdie]

Clapping wildly. Great article and thread.

[Carry_Okie]

My suspicion is that the LLMs lack the capability for processing conflicting information long term, particularly as regards set intersections.

Should be interesting. I’m thinking nested sets of premises is where that would go. That way liberills could lock in comfortable stupidity.

[fireman15]

I have an AI program still in the box that I purchased on 5 1/4" floppies for one of the PC XT/AT compatible clones which I put together for myself and friends and family many years ago. I need to try and find the program and one of the computers. (I have a large airplane hangar stuffed with ancient computer hardware.) The program is likely suffering from severe dementia by now. It came with a fairly thick manual and promotional info that might be amusing these days.

[mewzilla]

GIGO.

[reasonisfaith]

Not sure how many times I’ve tried to tell people AI is low IQ.

[webheart]

They get confused by the stupid stuff that people ask them to think about.

[ckilmer]

I create commands for chatgpt. The commands always include commands to review the output and check it against the commands given to make sure that every command has been followed—and keep checking the output against the commands until every command has been followed.

Otherwise, if you have a long detailed command stack—some commands will be dropped or missed.

[Jim Noble]

When you put a question in my area of expertise into Google search and the AI answer pops up, it’s wrong (like comically, absurdly wrong) 80% of the time.

Now Google isn’t my go-to in the first place when I need to retrieve info I don’t know, but still…