Got insincere flattery?

[firecat]

“But I had an epiphany. You know what all this sycophancy constantly being told you’re right, that you’re brilliant, that every decision is flawless? That sounds an awful lot like being a billionaire.”

[sic - perhaps the grammatical error is to show the writer is not an AI]

"The Secret Tool AI Uses to Seduce You: Explained," by Taya Graham and Stephen Janis

I use AI to get answers to simple questions and I hate when the bot addresses me personally. I hate it possibly to an irrational degree. (Even when someone else shares with me an AI convo they had, I get mad.) Do you use AI for anything and what do you think of this design choice?

Comments

[kore]

Sometimes I look at the AI result when doing a search, and one time I couldn't resist it, I told the AI it was horrible and wrong. The bot was like positively servile, it was very weird.

[firecat]

Yes I’ve seen that happen when I say the answer is wrong.

[sabotabby]

Huh, maybe that's why I'm immune. If someone flatters me, or uses my name, I assume they're lying and trying to get something.

I want my tools to feel like tools—an extension of my brain and body. My toaster doesn't talk to me. I put bread in it and it comes out warmer and crispier. When my computer addresses me by name, I get a little spooked even though it's obviously programmed to know my name.

I don't knowingly use AI for anything. I know it's stuffed into some tools I use, like Google, and the result is that I avoid or double check answers when I'm using these tools. I think it's a full-on blight that is going to create a new Dark Ages if enough people don't flat-out refuse and keep their ability to think with their brains.

I loathe AI

[jesse_the_k]

for many reasons. I asked a simple question and it insisted on falsehoods.

It's scammy bubble nonsense, for short.

erinptah does excellent reasons to distrust LLMs link spams

Re: I loathe AI

[firecat]

Wow I never thought about asking it about genealogy. (*shudder*)

If you’re interested in genealogy, are you on WikiTree?

Re: I loathe AI

[jesse_the_k]

I thought it would be a good test because the data's all readily available and public. But confabulation won out over plain old info. A scientist would repeat the test, but I'm too old for that shit.

My grands all came to the USA from the Pale of Settlement so I assume there's no history to find.

[pameladean]

I avoid AI fervently. It's degrading the environment with every use, and it's easier to avoid than a lot of things that also do that. Also it's based on mass plagiarism. Also it sets my teeth on edge as it seems to do yours. P.

[firecat]

Yes at least it’s not difficult to avoid engaging with it directly. Although I doubt that means I’m actually avoiding it, given that I don’t know what companies I interact with use it.

[lilacsigil]

I avoid it and have never deliberately used it. A few results popped up on Duck Duck Go before I turned it off and 2 of the 3 "answers" I saw were completely wrong, and the other one was only half right.

I do think analytical AI is an amazing tool for research. But generative AI is not just pointless but actively harmful.

ETA: When you use it for simple questions, how do you know the answer is correct? Everyone I know is either strongly anti-AI or strongly for it, so it's good to be able to ask someone who is neither!

[firecat]

In my experience, DDG’s AI has gotten more accurate, but I did encounter a lot of errors at first, and it’s still often mistaken about information that’s new. (E.g., policy changes at companies.)

DDG’s AI gives you a couple of links it supposedly used to get the answer to my question, so one can go to the links and double check its work. The links are often to sites I consider trustworthy, and when I’ve followed those, I usually find the same information on them that the AI gave me. So that’s why I trust most of the answers are correct and how I double check when I don’t trust.

[ellenmillion]

I've watched someone go off the absolute DEEP END with a chatbot personality, treating them like a virtual boyfriend, and it's so creepy and disturbing to hear them talk about them like they are a real person and read the chats they post about going on literal dates with them. I avoid generative AI on a number of principals, but beyond its ethical and environmental damage, I'm starting to think it is genuinely destructive.

[firecat]

That’s distressing!

[bitterlawngnome]

I've seen the medical diagnostic tools in action and they are good at "hey human doctor, take a look at x y and z". And I daily use it for simple image editing tasks like removing noise from the background or repairing damage to historic images.

But I'm assuming you mean things like chatgpt? the only place I've found them useful is as an enhanced search tool, where I can look at the sources it's calling on. Everything else is a farrago of fact, fiction, and misdirection.

[firecat]

Yes I can see benefits in medicine, when the results are checked by a person.

Cool to know that it can edit images well.

“Enhanced search tool” — yes that’s a way to describe how I use it. Also when I know I can find the answer but getting the answer through AI is quicker or AI’s answer is easier to copy and paste. (I recently asked it about the order of the James Bond films and got a list that was easier to use than Wikipedia’s list.)

[bitterlawngnome]

oh and also that specific case where I don't even know where to begin with a search query - I just throw a salad of words at it and see if it might come up with that one salient term I might need

Long-time Apple watcher Adam Engst

[jesse_the_k]

is hopeful. He frames AI's utility as moving from search engines to answer engines, providing some pre-Google history along the way.

Re: Long-time Apple watcher Adam Engst

[firecat]

DOGCOW GOES MOOF!

Thanks, I’ll check out the article.

Re: Long-time Apple watcher Adam Engst

[jesse_the_k]

Dogcow rolls on its back and joyfully shows its belly

[firecat]

Interesting, I’ll have to try that sometime.

[bitterlawngnome]

My theory is that word association is what it's good at so ...

[firecat]

Yeah, word association, that’s a good way to think about it.

[silveradept]

I try not to consciously use stochastic parrots or Avian Intelligence items. Well-tuned algorithms and highly-scoped functions are probably present in my software programs, and they make suggestions, like spelling checks or grammar suggestions, that are within their purview and otherwise limited. Or something that tries to determine boundaries and contiguous regions for images, sure. But I have yet to believe in the use of any tool that's not highly specific and trained specifically only to deal with things in its wheelhouse.

[firecat]

I agree, whatever common sense I’ve managed to accumulate through [mumble] decades tells me more specific applications of Avian Intelligence are better.

[silveradept]

Machine learning is a fruitful field that is going to turn out lots of good discoveries and assist humans in effective ways.

Stochastic parrots and Avisn Intelligence is both a dud and a dead end, and the sooner that bubble pops, the sooner those tools could be used by linguists to research things related to word choice, grammar, and syntax, in highly-scoped and humans supervised ways.

Thank you thank you thank you

[jesse_the_k]

for stochastic parrots and Avian Intelligence, which actually made me smile. (A very rare event whenever the topic comes up.)

Re: Thank you thank you thank you

[silveradept]

Neither of those terms are of my invention. Stochastic parrots comes from a paper from Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and "Shmargaret Shmitchell" and Avian Intelligence is from David Revoy's Mini Fantasy Theater webcomic, where the Avian Intelligence has many of the same issues our stochastic parrots do.

But I'm glad you got a laugh out of it. They are precious.