Right now the hype from most is finding issues with chatgpt

hype noun (1)

publicity

especially : promotional publicity of an extravagant or contrived kind

You're abusing the meaning of "hype" in order to make the two sides appear the same, because you do understand that "hype" really describes the pro-AI discourse much better.

It did find the fallacies based on what it was asked to do.

It didn't. Put the text of your comment back into GPT and tell it to argue why the fallacies are misidentified.

You act like this is fire and forget.

But you did fire and forget it. I don't even think you read the output yourself.

First I wanted to be honest with the output and not modify it.

Or maybe you were just lazy?

Personally I'm starting to find these copy-pasted AI responses to be insulting. It has the "let me Google that for you" sort of smugness around it. I can put in the text in ChatGPT myself and get the same shitty output, you know. If you can't be bothered to improve it, then there's absolutely no point in pasting it.

Given what this output gave me, I can easily keep working this to get better and better arguments.

That doesn't sound terribly efficient. Polishing a turd, as they say. These great successes of AI are never actually visible or demonstrated, they're always put off - the tech isn't quite there yet, but it's just around the corner, just you wait, just one more round of asking the AI to elaborate, just one more round of polishing the turd, just a bit more faith on the unbelievers' part...

I just feel like you can’t honestly tell me that within 10 seconds having that summary is not beneficial.

Oh sure I can tell you that, assuming that your argumentative goals are remotely honest and you're not just posting stupid AI-generated criticism to waste my time. You didn't even notice one banal way in which AI misinterpreted my comment (I didn't say SMBC is bad), and you'd probably just accept that misreading in your own supposed rewrite of the text. Misleading summaries that you have to spend additional time and effort double checking for these subtle or not so subtle failures are NOT beneficial.

I think because it's language.

There's a famous quote from Charles Babbage when he presented his difference engine (gear based calculator) and someone asking "if you put in the wrong figures, will the correct ones be output" and Babbage not understanding how someone can so thoroughly misunderstand that the machine is, just a machine.

People are people, the main thing that's changed since the Cuneiform copper customer complaint is our materials science and networking ability. Most things that people interact with every day, most people just assume work like it appears to on the surface.

And nothing other than a person can do math problems or talk back to you. So people assume that means intelligence.

While both Markov models and LLMs forget information outside their window, that’s where the similarity ends. A Markov model relies on fixed transition probabilities and treats the past as a chain of discrete states. An LLM evaluates every token in relation to every other using learned, high-dimensional attention patterns that shift dynamically based on meaning, position, and structure.

Changing one word in the input can shift the model’s output dramatically by altering how attention layers interpret relationships across the entire sequence. It’s a fundamentally richer computation that captures syntax, semantics, and even task intent, which a Markov chain cannot model regardless of how much context it sees.

Most humans don't reason. They just parrot shit too. The design is very human.

You either an llm, or don't know how your brain works.

Yeah, well there are a ton of people literally falling into psychosis, led by LLMs. So it’s unfortunately not that many people that already knew it.

LLMs (at least in their current form) are proper neural networks.

Fucking obviously. Until Data's positronic brains becomes reality, AI is not actual intelligence.

AI is not A I. I should make that a tshirt.

"if you put in the wrong figures, will the correct ones be output"

To be fair, an 1840 “computer” might be able to tell there was something wrong with the figures and ask about it or even correct them herself.

Babbage was being a bit obtuse there; people weren't familiar with computing machines yet. Computer was a job, and computers were expected to be fairly intelligent.

In fact I'd say that if anything this question shows that the questioner understood enough about the new machine to realise it was not the same as they understood a computer to be, and lacked many of their abilities, and was just looking for Babbage to confirm their suspicions.

LOOK MAA I AM ON FRONT PAGE

LOOK MAA I AM ON FRONT PAGE

Most humans don't reason. They just parrot shit too. The design is very human.

NOOOOOOOOO

SHIIIIIIIIIITT

SHEEERRRLOOOOOOCK

Extept for Siri, right? Lol

It’s an expensive carbon spewing parrot.

Fair, but the same is true of me. I don't actually "reason"; I just have a set of algorithms memorized by which I propose a pattern that seems like it might match the situation, then a different pattern by which I break the situation down into smaller components and then apply patterns to those components. I keep the process up for a while. If I find a "nasty logic error" pattern match at some point in the process, I "know" I've found a "flaw in the argument" or "bug in the design".

But there's no from-first-principles method by which I developed all these patterns; it's just things that have survived the test of time when other patterns have failed me.

I don't think people are underestimating the power of LLMs to think; I just think people are overestimating the power of humans to do anything other than language prediction and sensory pattern prediction.

LOOK MAA I AM ON FRONT PAGE

an llm also works on fixed transition probabilities. All the training is done during the generation of the weights, which are the compressed state transition table. After that, it's just a regular old markov chain. I don't know why you seem so fixated on getting different output if you provide different input (as I said, each token generated is a separate independent invocation of the llm with a different input). That is true of most computer programs.

It's just an implementation detail. The markov chains we are used to has a very short context, due to combinatorial explosion when generating the state transition table. With llms, we can use a much much longer context. Put that context in, it runs through the completely immutable model, and out comes a probability distribution. Any calculations done during the calculation of this probability distribution is then discarded, the chosen token added to the context, and the program is run again with zero prior knowledge of any reasoning about the token it just generated. It's a seperate execution with absolutely nothing shared between them, so there can't be any "adapting" going on

LOOK MAA I AM ON FRONT PAGE