Sure, and then we run software on them that does a variety of things besides math. Slack doesn't add numbers for me but it's incredibly useful anyway. LLMs aren't computers, they're software (more or less). I'm not an AI booster by any means but this kind of criticism is just silly.
Code is logic and word signifiers, not numerical calculations. The whole point of an LLM is that it converts complete words into numerical tokens. It’s not a calculator.
And LLMs aren’t useful for programming? Guess that’s news to every programmer then. That’s THE thing they’re most useful for.
Google will blatantly mislead you sometimes too, doesn’t mean it’s a useless technology.
Part of tech literacy is being able to think critically about the information you receive, wherever it comes from. No source is worth blindly trusting.
TBF; this is just the wrong tool for it. You don't ask your barber about the Feigenbaum constant and you don't use chat-gpt for bigger numbers equatioins. Of course current AI can handle it, but you need to train it for that.
Oh, I get it. The problem is education, as usual. LLMs are perfectly good at use cases that they were designed for. We have calculators and other tools (eg Wolfram Alpha) to do math.
They are good for programming. Particularly, code analysis, documentation generation, documentation review, conceptual questions, unit test generation, boilerplate code generation, comment generation, and full-line autocomplete.
One shouldn’t ask them to do things that have a right/correct answer, there are other tools for those kinds of problems. And they work best in collaboration mode, where through a series of prompts and responses, an LLM can generate useful output.
they are very good at writing things as long as nobody needs what they are saying to be 1. accurate 2. interesting
if you are rigorous with their training materials they -can- be used to summarize documents but they all have the same glitch: they always try to make the answer to your question "yes"
Muting this. If you hate AI, follow me. If you hate AI but think I’m annoying on here, follow @westernkabuki.bsky.social because I don’t have access to it. Goodnight!
If the hammer magically misses the nails or the screwdriver mangles the head of every screw all the while killing a dozen polar bears every use, yeah I think I could hate them pretty easily.
Yes, we need to understand that no AI is a magical troth-machine. They can help us and the sky is the limit, but at this point, we still need to prove it manually. They can do complex things in no time and our problem is we often can't prove it that fast, if the mistake is notobvious.
It has never helped anyone do anything except give up their own capacity for thought. It exists only to make shit up and jerk off tech shareholders. Otherwise, I have an NFT bridge in the Metaverse to sell you.
Wrong.
1. Auto-searching digital libraries did not make anyone giving up thoughts, too. It just helped us skipping annoying steps. Ultimately we learn from it because we need to reconstruct them anyway in order to understand and prove them.
2. helped us folding enzymes, creating new materials.
The sky is not the limit, AI is good at a very small sliver of things involving summaries(which already exist in the form of book reviews, Wikipedia, and abstracts) and the collation of large data trends for GIS, etc.
On one side of the ledger is the planet and the entirety of the human experience. What do actually think is on the other side of the ledger? Break it down for me. And don’t say curing cancer because AI literally cannot add two numbers together and scientists are making huge advances on cancer.
I've got a cheap saw, because I can't afford a good one. And while there's no way I can prove this to a stranger on the Internet...I do in fact hate that saw. It's a pain to use. It makes terrible ragged cuts. I have seriously considered throwing it out and just whacking away with a machete instead.
So it is *possible* for a human being to hate a tool. It's too early in the morning here for me to draw any kind of detailed analogy between cheap saws and LLMs, but the rest of the Internet can have at it.
look, if you haven't sworn a vendetta against the heirs of the Black and or Decker families at least once then have you really completed a home repair project?
Fixing my washing machine a few weeks ago I said something like, "who the fuck designed this? Engineers, I bet, never have to use the the things they make." Maybe not a tool but mad at the whole damn thing
This is such a dumb own. I own many tools in my fabrication workshop and have been a software engineer too. This isn’t some soft-hands take. It’s an actually nuanced and curious take that understands the technology in a way your black and white thinking will never let you.
If a hammer were designed in such a way as that it needed to first be bathed in the blood of a newly-slain infant before it could be used to drive a nail into a piece of wood, one might hate the fact that people were using that tool.
If i swing a hammer at a nail, but every time i do, it burns down a forrest, causes a drought and misses the nail then i can in fact call the tool shit and say it never needed to exist in the first place.
if my hammer has no weighted head, or one made of brittle plastic; and my screwdriver has a round, smooth tip, yet are presented/advertised to me as groundbreaking tools despite being incapable of doing simple jobs AND are made via exploitation and accelerate environmental destruction, i can be mad
Of course you can hate a hammer or a screwdriver if it is malfunctional, expensive. You can hate every tool if you will. I Hate Windows, I hate the Microsoft office, and I hate chatGPT and all genAI unethical applications
There actually is a hell specifically made for tech dipshits who think using the power equivalent of a household to wrongly answer a question every time you Google search is a good thing.
Yes, I hate them too. No, I'm not one of them. I am currently working on a new architecture that helps fixing this. At least atm it seems like standard AI can't get better with more upscaling. And that's a good thing.
Sure I can. Someone sells me a screwdriver made out of some crap alloy with a slightly wrong-sized head, I'll hate the hell out of it, and the guy who made it.
What if that hammer sets your house on fire because other people looked at it? Or it consistently bent nails no matter how perfect your technique? Can we hate it then?
Imagine an electric screwdriver that requires a direct connection to a nuclear power plant and a river diverted to cool it, and it only sometimes turns the screw in the right direction.
It purports to do math. It appears to do math. It silently gives wrong answers for math, and many other questions. It’s unfit for most of the purposes it’s touted for.
It's a tool to make rich people more money, it doesn't do anything useful or better than what we already have. It largely just sucks, but as a parlour trick it's impressive, so stupid people are very taken with it.
Environmental vandalism, grifting venture tech investors that shouldn't be trusted with the money they manage, briefly interesting people who were briefly interested in ELIZA. You could use one instead of Lorem Ipsum if you needed it to technically not be Lorem Ipsum for some reason?
If a hammer didn't do what it was intended to do 90% of the time and there were already other tools available that did the same thing more accurately I would absolutely hate hammers. "It's a tool" isn't an excuse. It's a BAD tool that doesn't fucking work.
My guy has never used a cast ball joint splitter and had it fail by not fitting where needed or it exploding on him before actually doing the task it was made for.
AI is a tool, true.
Generative AI, however, isn't. It's a parasite poisonous even to itself.
If you had never seen a painting and had a camera, you could take a picture still, but without stealing images you could not produce one through generative AI.
i hate screwdrivers. they're so annoying, but i gotta use em. your analogy doesn't make sense, as unlike AI, my hammer doesn't make up nails or suck up unholy amounts of water that could be put to better use
As the last starving polar bear drowns in an iceless sea, I reach for my hammer but it’s actually a stick of rancid butter which is also someone really racist
A hammer doesn’t steal and siphon other creators content without their permission, nor can it be used to trick elderly over the phone. A hammer doesn’t track my where about and locations to sell to advertisers.
AI ultimately won't also. these huge energy resources are used for training. It needs these servers to get the packed model, but after that, you can run them locally and it's getting better. luckily, more servers does not scale everything and current AI reached that point.
What if - and I’m just spitballing here - we hate on the companies actually stealing people’s work, burning down the environment, and trying to put people out of jobs?
AI that doesn’t do those things is *completely achievable*! But only if pressure (or better yet: regulation) makes if happen
I can be mad at a hammer that breaks if I try to hit a nail with it and is also somehow using more energy to do so than the entire city of Toledo, Ohio.
I was looking for a shortcut summary to explain standard deviation and average range one day and the AI popped up in the search with the wrong answer and I didn’t even look further just wrote my own,
lol I know and I’m sure AI can do amazing things but when i use chat gpt I’m not super impressed unless I’m Using it to write procedures that’s pretty good. But otherwise I can google that stuff
I once tried to apply a math problem from a FoxTrot comic strip where an answering machine message tells the caller to press a number, but phrases it in the form of a long math equation. The correct answer was of course 1, but AI said it was a long negative number that doesn't exist on a dialpad.
I'd show you the original strip, but it seems that just a few weeks ago GoComics rebooted and switched to a paid subscription model and no longer allows guests and free users to access unlimited archives past a few weeks from today or something.
I DO know it's from the last decade of dailies tho.
Sorry: a calculator that costs 1 trillion dollar, uses unpaid labor from the global south, burns a forests and drains a lake for a simple calculation for no reason at all, AND THEN gets it wrong.
Känns som att det här borde vara en enkel sak att lösa. Om ChatGPT känner igen att det rör sig om en matematisk beräkning så kan den väl slänga in inputen i excel eller något?
Well, the salesman is selling a motorcycle as if it’s a car, with AC, custom paint and self driving capacity.
It gets half a mile a gallon, the “custom paint” is a copy of another bike, the exhaust pipes don’t actually seem to connect to anything, and it will occasionally steer into walls.
I once spent 20 minutes debating with ChatGPT about numbers. That’s her weakness. Just needed her to count the number of urls listed in a document. Yet, each time I asked her to check again, she had a different number. Every time. I asked why a 5 yr old could do what she couldn’t & she agreed. Weak.
Ooh, good, but like, what if the calculator randomly was sometimes right and sometimes wrong and the only way to know which was to recheck everything by hand yourself, anyway? Maybe right often enough to encourage people to think it wasn't worth bothering to check, though?
Comments
So, it gets math wrong.
"They exhausted every resource on the planet to make over one trillion of them, in order to make one that actually worked."
Me when I use a hairdryer as a calculator and it gets the answer wrong: 😡😡😡🤬🤬🤬🤬
And LLMs aren’t useful for programming? Guess that’s news to every programmer then. That’s THE thing they’re most useful for.
Part of tech literacy is being able to think critically about the information you receive, wherever it comes from. No source is worth blindly trusting.
BENDER: I need a calculator!
FRY: You *are* a calculator.
BENDER: I mean a *good* calculator!
It’s not wrong, most people will also get a math question wrong.
The question is why we felt we had such lack of bullshit that we needed to industrialize bullshitting.
Why would it not be good at math?
Of course you and I know that the overhyped autosuggestion machine isn’t actually AI, but you get why people would conflate the two, right?
Because unless you know what you’re doing so you can correct them, they get things wrong just a little too much.
if you are rigorous with their training materials they -can- be used to summarize documents but they all have the same glitch: they always try to make the answer to your question "yes"
I call it "The mathematically weighted plausibility machine." 😂
Idiot: Sounds good to me!
1. Auto-searching digital libraries did not make anyone giving up thoughts, too. It just helped us skipping annoying steps. Ultimately we learn from it because we need to reconstruct them anyway in order to understand and prove them.
2. helped us folding enzymes, creating new materials.
The problem is that it’s advertised as such.
*That’s* what gets people’s backs up.
Skill issue
Get better at hating
I sure as fuck can hate a tool
Sheeeeiiiitttt......
LLMs hallucinate *by their fundamental design*. it's giant autocorrect.
Look, you're already here and you use a fraction of the water.
Fork separator gang 4 lyfe
Generative AI, however, isn't. It's a parasite poisonous even to itself.
If you had never seen a painting and had a camera, you could take a picture still, but without stealing images you could not produce one through generative AI.
AI that doesn’t do those things is *completely achievable*! But only if pressure (or better yet: regulation) makes if happen
LLMs are not.
https://bsky.app/profile/r5-to-philly.bsky.social/post/3lcngppouz22a
It's not a tool, it's a scam.
why would "genAI" be excluded from that hate while it makes shit up, steals from actual human work and burns immense energy & water
*rips away curtain*
Oooo... Oh no... Don't look at that... Um
I DO know it's from the last decade of dailies tho.
Seems dumb, but important. It can't actually understand the request
It's also why it will argue things that are plainly untrue, because it's not reading, It's interpreting and processing.
Basically, it's a pretty solid ESL speaker.
Of course, this is going to bite them in the ass, because info from AI will ruin their own decision making processes.
I hand counted and asked if it was sure. It said yes.
I told it my result and it said “ope you’re right, sorry about that!”
It gets half a mile a gallon, the “custom paint” is a copy of another bike, the exhaust pipes don’t actually seem to connect to anything, and it will occasionally steer into walls.