A note from one of my favourite authors at the moment.
@thirstybear Also, fire does have constructive uses. Not so LLMs.

@jmax

@thirstybear I think they mean machine learning, not llms

@yugthebug @jmax @thirstybear

LLMs are like AN/FO explosive when compared to most uses of machine learning style fire.

"You object to me blowing up your house but I see you using a gas stove."

@resuna @yugthebug @jmax @thirstybear You probably use an ML system every time you spend money - every transaction is fraud scored at some point, some multiple times.
@TimWardCam
*Someone* (bank or payment processor) uses ML (edit: which are not LLMs) to check my transactions, not me and not (usually) the party I'm actively engaging with. What was your point?
@resuna @yugthebug @jmax @thirstybear
@ScriptFanix @resuna @yugthebug @jmax @thirstybear That not all "AI" is LLM and not all non-LLM "AI" is evil. Many people here get that, but there are plenty who don't.

@TimWardCam @ScriptFanix @resuna @yugthebug @jmax @thirstybear

Oh come on, when people shit on AI they are not coming for camera autofocus.

This is a moral issue, directly linked to active genocide, and most AI true believers publicly hope for human extinction.

@TimWardCam @ScriptFanix @resuna @yugthebug @jmax @thirstybear people know that. That’s why “AI” is in quotations.

@TimWardCam

Are you uh... Trying to prove the original post? 'Cause you're talking about gas stoves here.

@ScriptFanix @TimWardCam @resuna @yugthebug @thirstybear If you use the term AI in modern language, you're talking about LLMs. Part of the scam is to pretend that they are synonyms, and they've won that fight.

You're helping with the con by continuing to refer to actual useful technologies as "AI".

Scalzi meant AI in the colloquial sense of LLMs. Stop playing "well actually" games that only benefit con artists.

@jmax 💯
They (LLM pushers) successfully took over the meaning of "AI". Let us all accept that and use more specific terms when we mean something else. AI has never been a well defined term anyway and can be used to describe about anything
@TimWardCam @resuna @yugthebug @thirstybear
@ScriptFanix @jmax @resuna @yugthebug @thirstybear The one that pisses me off is "organic", but I accept that that one's lost.
@TimWardCam @ScriptFanix @jmax @resuna @yugthebug @thirstybear Well actually... *scnr* in computer games I shot AI – for the past two to three decades. It's fun. And in recent games, e.g. Arc Raiders, AI behavior is actually machine learnt. Even funnier to outsmart that...

@TimWardCam @ScriptFanix @jmax @yugthebug @thirstybear

Organic, chemical free. I guess it's some kind of crystallized energy.

@ScriptFanix @jmax @TimWardCam @yugthebug @thirstybear

Don't refer to any of them as AI, just because they're lying doesn't mean you have to lie too.

@LeBonk @ScriptFanix @jmax @TimWardCam @yugthebug @thirstybear

I just call them LLMs. AI is a field of research. LLMs are a spin-off, but they're not AI in the science fiction sense anymore than a Fisher Space Pen is a rocket just because it's a spin-off of the space industry.

@resuna @ScriptFanix @jmax @TimWardCam @yugthebug @thirstybear
yeah but this also includes image generators and things that maybe can't be accurately described as dealing with "language" in its output
Kate Crawford's term "large-scale computing" is also good as a reminder that these things are really just the same sort of thing we've had for a while, but bigger, more resource-intensive, and more exploitative
@resuna
@ScriptFanix @jmax @TimWardCam @yugthebug @thirstybear
also if you wanna be more specific, u can say something like "plausible sentence generator", "plausible image generator", "plausible code generator", or whatever
so, for example, Minnesota recently banned plausible nude generators
@ScriptFanix @jmax @TimWardCam @resuna @yugthebug @thirstybear AI is a meaningless term. For a start, nobody really believes that machines, algorithms or tools are really sentient beings (as far as I can see, the only ones who claim that are the notorious liars who actually make the tools and a few people who are of sub-par intelligence who have been duped by their marketing), or in any meaningful sense, 'intelligent'. So, we shouldn't use the term at all. Call them slopbots, slop distributors.
@UkeleleEric @ScriptFanix @TimWardCam @resuna @yugthebug @thirstybear It used to be useful as a catch-all term for a loose collection of academic research on "things it's hard to program a computer to do". No more.

@jmax

@ScriptFanix @TimWardCam @resuna @thirstybear yeah I get that ai refers to generative ai nowadays, which is why I clarified to machine learning which does in my opinion have legitimate use cases

@TimWardCam Lots of comments arguing that this is not what people are complaining about, so I felt I should point out whatever system my cash back for groceries card is using for fraud detection is an absolute sack of shit, excelling at both Type 1 *and* Type 2 errors.

@resuna @yugthebug @jmax @thirstybear

@yugthebug @thirstybear No. There are no commercial services that use anything except LLMs. The techbros really belive it's a universal solution.
In the modern world, AI == LLM.

@jmax @yugthebug @thirstybear Weather Prediction? Valve Anti Cheat (since at least 2019)?

They're out there, they're just not screaming it into the void.

@jmax @yugthebug @thirstybear I was only refuting the claim about no commercial services using non LLM AI.

Everything else that was said is true

@krutonium

Just to add some lol to this: Australia's BOM are using Anthropic's LLMs to "assist" the accuracy of their predictions.

@shaknais that's... Not *really* what I was talking about lol.

Modern Weather Prediction Models are basically trained models that don't communicate in a normal language, but in weather.

You feed them historical data, current weather, and other relevant data like the date, and it predicts what weather comes next - and actually more accurately and faster than traditional weather modeling.

@krutonium

Yeah, I understand the modelling. I just don't understand the BOM who had all that, and decided to throw an LLM on top of it, and then got confused when their accuracy dropped off so sharply there's a Senate inquiry.

They're using Claude. Just Claude.

@LevZadov @yugthebug @thirstybear I'm sorry, but you've lost that fight. In general usage, LLM is exactly what AI means.
@thirstybear Did you know there are legit use cases for stabbing strangers and taking their money. For example, if I were to run a tattoo shop, I wou- what? You need an ambulance? I am not convinced you are losing a dangerous amount of bloo- ...sir! Please be civil! I am trying to engage you in a constructive debate!
@thirstybear Ban fire, of course!

@thirstybear Ah, thanks for sharing it. I found the quote on Bsky and shared it over there.

https://bsky.app/profile/scalzi.com/post/3mj6q3pdblc2y

John Scalzi (@scalzi.com)

I don't know how to explain to the people who pop up when creatives complain about "AI" to say there are legit uses for it just how much they sound like someone saying "Well actually fire is used to make bread" when people are talking about an organized arson ring burning down their fucking houses

Bluesky Social
@thirstybear I enjoyed his book about the accidental supervillain! What else of his should I read next?
@dneary
Redshirts, The Kaiju Preservation Society
@thirstybear
@DelilahTech @dneary Another vote for Kaiju Preservation Society here. I’m new to the author so playing catchup, but everyone mentions the Old Man’s War series.
@thirstybear
It's a good series, I just find his standalones more entertaining
@dneary
@dneary @thirstybear The Old Man's War series is really good.
@dneary @thirstybear the Old Man's War series is pretty good
@thirstybear in my long quest to find AI any use, I found "one shot usage programs". So you generate them once. They do a thing by only running once, and then you never ever need them again. Especially you don't need to maintain them as you delete the program immediately after use.
That's a very specific use case. Security researchers do that.
@thirstybear - and this assunes there is, in fact, legitimate uses for AI, which very muchremains to be proven...
JWcph, Radicalized By Decency It’s been very good foor the rapid spread if disinformation!
@jwcph @thirstybear I think that depends on if it means “machine learning as a field” or “specifically LLMs and diffusion models”
@thirstybear One problem with this analogy is that the people talking about the arson ring often suggest banning fire rather than stopping arsonists from misusing it and the only way to get them to focus on the arsonists starts with reminding them that fire can be useful.

@MartyFouts or, and hear me out here, you could put your whataboutism where the sun does not shine. no-one loves proselytisers.

@thirstybear

@mawhrin @thirstybear oh aren’t you a special snowflake. Have a nice day.
@mawhrin
I blocked him months ago for this behavior.
https://tech.lgbt/@toolbear/116490307172287700
toolbear#🌶️ (@[email protected])

I recently learned that pedantry around what "AI" means is an example of Presciptivism. https://en.wikipedia.org/wiki/Linguistic_prescription While I'm generally in favor of precise language, there's a style of bad-faith, pedantic reply about using "AI" instead of "LLM" or whatnot. That's Linguistic Prescription (pejorative) in action. #AI #LinguisticPrescription #Prescriptivism [this is, in part, a subtoot about Marty Fouts]

LGBTQIA+ and Tech

@MartyFouts @thirstybear they want to ban their houses being used for fueling the fire. The people roasting marshmallows and baking bread on their burning houses are not entitled to their houses as fuel and the arsonists have convinced them it’s necessary and okay.

So no, that’s not a problem. You just forgot that part of the metaphor.

@707Kat @thirstybear Well you have certainly belabored the metaphor more than I thought possible but your example still has people going after the wrong people, your “marshmallow toasters” instead of the arsonists.