Google AI just plain wrong...

[basscooker]

.. on a "kinda insignificant" fact; however still wrong.

I'm not an enemy or trying to "gotcha", but how would one go about correcting or teaching their AI they got something wrong?

 

[porkynuggets]

AI is going to be wrong most of the time, it's learning from the average output of humans

 

[Winslow]

"AI" as it currently exists is not even close to the perfectly-correct vision of Star Trek, etc. Right now, it's mostly just venture capital hype, with a primary feature of separating money from idiots ~ahem~ "investors" who don't read the fine print, or concern themselves with reality.

AI is going to be wrong most of the time, it's learning from the average output of humans

Exactly.

Garbage in, garbage out - Wikipedia

en.wikipedia.org

 

[MonetBass]

AI is only as good as the database behind it, and right now, most of those are not good.

 

[Mike Vee]

the Google one really sucks. Try ChatGPT

 

[skycruiser]

Ignore it. Skip past it in your search results. (But in general people are lazy and won't skip it.)

 

[Major Seven]

"AI" as it currently exists is not even close to the perfectly-correct vision of Star Trek, etc. Right now, it's mostly just venture capital hype, with a primary feature of separating money from idiots ~ahem~ "investors" who don't read the fine print, or concern themselves with reality.


Exactly.

Garbage in, garbage out - Wikipedia

en.wikipedia.org

and that includes bias, prejudice and pervasive urban myths.

 

[Micha84]

.. on a "kinda insignificant" fact; however still wrong.

I'm not an enemy or trying to "gotcha", but how would one go about correcting or teaching their AI they got something wrong?

You could simply tell it. If the programing is not completely useless it should consider the input.
Speaking about chances of success, however... may as well send a letter to Apple calling their iPhones stupid.

 

[DrBnz]

Well think about it. AI is generated by the inputs of the masses. Now look around you. See what I mean? Them folks aren't exactly the people I'd be taking advice from.

 

[ProbablyTooLoud]

I do some pickup work training LLMs.

If an AI makes obvious factual mistakes it shouldn't be in use.

And let's be clear that this is not AI as old scifi heads understand the term.

The singularity has not happened.

 

[Major Seven]

Well think about it. AI is generated by the inputs of the masses. Now look around you. See what I mean? Them folks aren't exactly the people I'd be taking advice from.

This is my point. It is only a reflection of us: popular untuths and falsehoods included. The police crime software even "invented" profiling by itself.

 

[Major Seven]

Well think about it. AI is generated by the inputs of the masses. Now look around you. See what I mean? Them folks aren't exactly the people I'd be taking advice from.

The wall mart paradigm.

 

[MuseChaser]

..

And let's be clear that this is not AI as old scifi heads understand the term.

The singularity has not happened.

I don't think it was just us "old scifi heads." Artificial Intelligence, even in the cutting edge scientific community, used to indicate a level of self-awareness and creative thought in a machine. The terms have been corrupted, like so many other words whose meanings have been changed. My youngest son is fond of saying, "Language evolves" and "we all create our own reality." From my view, it's more like "language becomes corrupted" and "we are free to live within our own delusions, but reality just IS." Who knows who's correct.

 

[Pilgrim]

AI tends to be right in general terms, but may be wrong and occasionally invents details and quotations.

AI is basically scraping the written words which are accessible on the web and compiling a report from them. Of course, the more scientific and NON-politicized material it complies from, the more accurate it is.

At Ronald Ray-Guns said, "Trust, but verify."

 

[shadowtippy]

Well think about it. AI is generated by the inputs of the masses. Now look around you. See what I mean? Them folks aren't exactly the people I'd be taking advice from.

True. There are as many people with below average intelligence as there are with above average intelligence.

 

[S-Bigbottom]

If it pulls its data from this page, then it knows it sucks.

 

[Ricky Rioli]

I was having a nice chat with Google AI about artists, and we were trying to find me some lesser known contemporary artists to enjoy .....

Which sounded intriguing, but.....

 

[gln1955]

AI seems to be programmed with built-in Dunning Kruger. The more wrong it is, the more confidently it presents its "facts".

 

[Ricky Rioli]

If you chat to Google AI about facts and language, it cheerfully admits that it knows nothing except which words like going together and how: the facts come along by accident, as it were. And if the words that seem to want to go together don't relate to reality, oops! An articulate database it most definitely is not.

PS I'm still annoyed that that painter does not exist: Google did an excellent job of working out what words most wanted to come next in our conversation — unfortunately this required leaving the known world behind.