Chillbo Hangins: Google explaining the “glue pizza” search result by stating EXPLICITLY that the wrong answer was given BECAUSE nobody had ever asked that question in the training data tells you that what we have isn’t artificial intelligence. It’s a server that copies, changes slightly, and pastes. Not intelligence

Google explaining the “glue pizza” search result by stating EXPLICITLY that the wrong answer was given BECAUSE nobody had ever asked that question in the training data tells you that what we have isn’t artificial intelligence. It’s a server that copies, changes slightly, and pastes. Not intelligence

Okay this is looking like a banger so it’s: free Palestine from the river to the sea. Period.

Hundo “AI” is a (energy intensive) copy-paste machine

That copy-pastes wrong a lot of the time.

Ted Chiang calls AI an "unreliable photocopier."

NFTs all over again.

And blockchain before that And crypto before that … … And nanotechnology before that And…

Don't you mean copy-pizza sauce?

And Imma keep beating this poor dead equine: If the so-called AI isn't capable of parsing, understanding, and integrating petabytes of data to produce insights a human could not (due to the volume of information more than intelligence), what the fuck GOOD is it? 1/2

2/2 I used to link to this 1948 story because it predicted the modern Internet better than 1980s cyberpunk did, but lately, I link to it because it shows what an AI that could actually draw *new* conclusions from data might do. And we are so far from that. www.baen.com/chapters/W20...

- Chapter 2www.baen.com

What a delightful story, thank you

ML is good for finding patterns... It's how modern astronomy is done. Just... Stars have more predictable rules than humans.

like this is crucial: these AI models are okay at "make thing look like thing" but they have no capacity to understand "WHY thing look like thing" they have no capacity to understand anything at all everything they produce is noise which happens to LOOK like signal

i'm not being hyperbolic when i say that the chief proponents of AI nonsense are intent on waging war against meaning itself when Sam Altman asserts that we are all "stochastic parrots," he is asserting that AI's lack of understanding is not a problem, because "understanding" itself is a fiction

Right. That’s the essence of the “seed”. It’s the index of the record that assembled the most acceptable approximation of appropriate noise. There’s no “experience” or “rationale” behind the choice. It would only be intelligence if intelligence was based on an absolute mathematical calculation

That's a great metaphor.

and if you argue it is better at doing your job than you are (which I got from an AI proponent this morning), how good are you at your job?

It makes you wonder ‘then why even bother to change things slightly?’ and that’s when you click it’s 100% about intellectual property laundering — and that’s why it’s sociopathic tech billionaires who are so bizarrely keen to put AI in everything.

Intelligence is being able to parse the input "can you put glue on pizza" and produce a coherent response that sounds like it was written by a human. Wisdom is knowing that you don't put glue on a fucking pizza.

Charisma is being able to actually sell this nonsense to folks with a straight face.

Endurance is eating the pizza

That should have been Constitution, to keep it within the same ruleset 😆

Strength is stopping anyone trying to keep you from eating the pizza

Sexually Suggestive

Labeled by Bluesky Moderation Service

Dexterity is dodging the piece that someone’s trying to share with you

It's a dumb plagiarism machine.

It's not AI, just a more sophisticated chatbot.

A regurgitating cow that spews out everything, especially bullshit.

It's a compression algo that runs on megawatts.

Yes, our normal (pre-cybernetic) use of the word “intelligent” (-ence) insisted on “good judgement,” “sound thought” and “rationality/reasoning” all of which are conspicuous by their absence in AI parroting

Yeah, It didn't give that response because it wasn't in the training data questions, it gave that answer because currently, LLM AI is incapable of understanding satire—or just outright lying—that it comes across. By definition, it doesn't understand WHAT it is looking for, just HOW to look for it.

Yes. In essence it’s incapable of original thought or rationale. All it does is compare with existing knowledge regardless of accuracy of relevance. Oh! And actual intelligent beings can say “I don’t know”. Or “I can’t find the correct answer”, but because capitalism cannot allow that, we get bunk

yes. it's almost like it's some kind of transformer. some kind of generative pre-trained transformer.

Knowing what a large language model does tells you that what we have isn't artificial intelligence.

And, once again, we should consider how “nobody thought to ask the question in training the data” impacts underrepresented communities. Cultural nuance can’t be captured if the culture isn’t part of the process.

Qualified to replace most mid-level corporate jobs then?

Adult Content

Labeled by Bluesky Moderation Service

Haha!

The funny thing is that they have the compute power to ask ALL the questions ahead of time... It's like the indexing loses context or something...

Never was, never can be. It’s branding.

*and* drinks all our water

"If you want pizza without glue, give us a better recipe to plagiarize!"

I found this article to be a good explainer of the very basics of how AI works, and also why it fails to work so often. techcrunch.com/2024/06/01/w...

WTF is AI? | TechCrunchtechcrunch.com What is AI? We've put together this non-technical guide to give anyone a fighting chance to understand how and why today's AI works.

As I have said for 2 years, being an old programmer.

"It copies and pastes. And speaking of paste, guess what goes on pizza?"

Haha! 😆

LLMs are just guessing engines. They’re not intelligent.

Yeah. But my good friend, you get why one would think they’re supposed to be intelligent yeah? Because of the product description? In all their keynotes and junk?

Indeed

They’re not even LLMs, just synonym engines looking for something that’s similar and swapping words out.

Lies, damned lies, statistics, and AI

So Google is basically like all the retailers who install self-checkout kiosks and then employ a skeleton crew of minimum wage staff to stand around waiting for predictable problems to arise. “Who cares if it sucks for customers. They gotta [buy groceries/search the Internet] somewhere.”

Sounds a bit like they’ve managed to successfully virtualise the average user of certain social media - incapable of telling fact from fiction and regurgitating utter nonsense.

Post