Crackpot nonsense. Humans are not LLMs. This simple fact should be obvious to anyone with even a superficial understanding of LLMs.

Wow. You're still going to double-down on your obvious nonsense?

You're beyond hope.

Artificial? LOL! As opposed to a natural "assembly language interpreter". Forth is not an assembler, neither is it an "assembler interpreter", artificial or otherwise.

Forth is absolutely a high-level language. A surprisingly powerful high-level language at that.

You couldn't be more wrong.

You've previous claimed to have used Forth for years, yet you don't seem to know anything about it? You're either a liar, very stupid, or both.

So, no, I don't think that I can explain the basics of LLMs to you. You are incapable of learning, incapable of admitting error, and confident in your deep ignorance.

There is no confusion about what the letters stand for. Also the term "Large Language Model" does not tell you anything at all about what they do, how they function internally, or their limitations. What the fuck is wrong with you?

Because morons like you say stupid things about LLMs. I am, of course, assuming that people say things that are somewhat consistent with what they think. In your case, however, I only use the word 'think' for lack of a better term.

What argument? You said something irrelevant. I agreed that you don't understand my post and gave you some much needed advice.

If it doesn't hurt to be you, it should.

Take your own advice. You obviously don't have a clue. I certainly can't give you the education you need in a slashdot post, not that you're capable of understanding the material anyway. You have ... a history.

Have you ever been right about anything here? In every conflict with you I've seen, you've vehemently defended the most ridiculous assertions. Hell, you argued for days that Forth, a high-level programming language, was an assembler. I saved this one, because it was just so unfathomably stupid. It's one ridiculous thing after another, zero self-awareness. I actually use this as an example of what to watch out for in interviews.

The point, of course, is that explaining why all of that is laughably absurd takes a lot more effort than it took you to vomit that out. For example, let's look at just two lines in your manifesto of nonsense:

Just for fun, this can be done in less than 10 lines in BASIC, with nothing more than common simple operations. No "odd" or "esoteric" tricks required. It should take just a few minutes, not 30 years. I would expect any junior to be able to handle a task like that in under an hour. This would need to be followed by the code, to prove the point. You'd object to the language (selected to make you look even more foolish) so I'd be tempted to show how it can be done in C, though you'd complain about "esoteric tricks" like bitwise operators, so I'd need to show how it could done in Java, a verbose language in which you have often claimed expertise. Then there's the doubling-down and goal-post shifting that inevitably follows ... all for just two lines of your nonsense about a much simpler topic than LLMs.

Let's look at your insane claims here:

What possible basis do you have for making this claim? Why does having "random humans" writing prompts force "arbitrary or global variable names"? How do you distinguish between names that are arbitrary and those that are not? Why could they "only solve isolated tasks"? It's completely incoherent!

How is it obvious? What about being agents means they can "agree". What do you think that LLMs do? What do you think agreement means in this context? How are you defining "linkable" code? Why can agents do this but humans can not? What is your evidence to support this nonsense?

Your problem is that you think that LLMs are science fiction robots. They are not. They are obviously and objectively not, as anyone with even a superficial understanding could tell you. I could spend days trying to explain to you how they work, but you've proven time and again that you can't even handle the basics. You'll never be able to handle the math, even if you put in real effort, which you are unlikely to do. Ultimately, it doesn't matter what I write. Nothing I can demonstrate will ever change your mind. Your posting history is enough to prove that. It does not matter how simple your nonsense claim is, you will defend it in the face of overwhelming evidence to the contrary. On a complex topic like LLMs, there is no possible way I could convince you to abandon your nonsense beliefs and embrace reality.

That's pretty stupid, even for you.

That's because you're stupid.

You should take your own advice. You have no idea just how stupid your comment looks here.

I hear people blindly repeat that particular 'fact' pretty often, but it does not mean what people seem to think it means. I've also seen the claim stated as "neural networks are good at spotting patterns in data".

When people hear things like that, they imagine some hyper-intelligent science fiction robot spotting patterns in seemingly random data. They might even think that modern LLMs can be used to analyze large amounts of data and find hidden patterns. This is silly nonsense.

When we say that 'AI is good at finding patterns in data' there are two things we could mean: using a model or creating a model. When using a model to "find patterns" we're using to model to compare the data of interest to the data used to create the model. Image recognition is the obvious example here. Creating a model is where the real 'pattern finding' magic happens. Neural networks, for example, can be used to model a function with nothing more than examples of relevant input and output. Clustering can be used to find group data in high-dimensional spaces and even spot anomalous data.

For clarity, you can feed random data to a neural network (or any other model) all day long, but it is never going to magically 'spot patterns' ... or do anything meaningful at all, really. Despite bad pop-sci, AI are not science fiction robots and neural networks are nothing at all like brains. NNs only "learn" because we modify their weights based on how closely their output matches what we expected given a particular input. They're really just simple functions, after all. No matter the approach, any resulting AI models are just, well, models. The process of creating them is where all the action is.

No.

This "it's okay that AI makes mistakes because humans aren't perfect either" is ridiculous. The kinds of mistakes humans make are so far removed from the kinds of 'mistakes' LLMs make that it's misleading to even call them mistakes.

Humans, for example, will never accidentally summarize a document that don't exist.

LLMs are not science fiction robots. They are not 'electronic brains' that think and reason. The very idea is absurd on its face.

+5 Underrated

Your gaslighting isn't going to work, dipshit. We can read. We know you're completely full of shit.

You're such a fucking joke.

So... you can't actually support your bullshit claims.

Here's the thing, champ: Unlink the average right-wing moron, we actually read the linked content. We don't just blindly assume they support your bullshit.

You're such a fucking joke.

You're a dishonest scumbag. The article does not support your bullshit claims.

Fuck off, troll.

Are you implying that American citizens do not exist?

With the constitution in tatters, you might be right about that...

He's just not being clear. By "people would find an iDevice barely useful without" he really means "Please, don't use lockdown mode. It makes you harder to oppress. Privacy is treason."

No one deserves to fail because they guessed wrong and built some critical bit of their infrastructure on what would turn out to be an abandoned product. Neither does anyone deserve to fail because they can't manage to transition away inside of some arbitrary timeline you set based on zero information. Four years is really not that long, particularly for a critical migration.

Do you seriously think that just because the original developers used Flash that the software must necessarily be simple enough that it can be inexpensively scrapped, rewritten, and replaced? Do you seriously think that any migration from a system developed in Flash must necessarily be trivial? Just how limited is your experience?

Migrations are not nearly as simple as you seem to think. Just rewriting the software (if you can even get that far) is far from the only obstacle and it's often the easiest part of the whole process. I've even seen migrations from "mission-critical" spreadsheets fail just from internal resistance. It doesn't matter how slow, complicated, and tedious the old system is, or how much faster, simpler, and easier the new system is if you can't get the users through the transition. Even with everyone on board, there's still no guarantee that the transition will be successful.

You'd be shocked at just how many successful businesses depend on some antique bit of software, possibly also running on antiquated hardware, at considerable expense because the consequences of a (or another) failed migration are so high.

Do they all "deserve to fail" if they can't migrate to the latest fad inside four years? Honestly...