Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

How LLMs work

0xkato.xyz

Post reply on HN
Print view View original

281–290 of 293 posts

Re: How LLMs work

#281

Post by spaceisballer »

layla5alive wrote: post #253
spaceisballer wrote: post #232

Earlier quoted context omitted.

I have no doubt there are similar stories to yours or to a lesser extent. And while I’m glad you have gotten some relief, what price are we paying as a society? I’m not just talking about the environmental impact alone, what about the societal impact? Social media has not been around too long but it seems to be a net negative for society. I’m sure there are plenty of anecdotes to the contrary, but the studies are sho…

I think most of the nuance in your first argument was undermined and overshadowed by your fairly categorical assertion that LLMs are bullshit generators and not actually useful. When you got a reply illustrating how incorrect that claim from your first argument was, you shifted to focusing on the other argument (the one I actually happen to agree with - the cost to society of hitching increasing dependency on big tec…

I actually didn’t say that aren’t that useful. I would argue that I don’t think the payoff is worth the cost. The areas investing unholy amounts of money and passing it among the largest companies are seeing plenty of cash, me and in general the public will only experience the squeeze when the huge profits aren’t realized to offset the immeasurable amount of money “invested”.

Re: How LLMs work

#282

Post by locknitpicker »

Chu4eeno wrote: post #279
locknitpicker wrote: post #259

Earlier quoted context omitted.

> The big breakthrough is we can interact with the agents using natural language - because of the LLM. Without ReAct and tool calling, all you have is a chatbot. That's useful, but it's just a chatbot. ReAct loops and tool calling is what unblocks high value usecases. It enables systems to actually address free-form problem statements, gather data that is not a part of their training set, inspect the current state of…

It's not very novel, though, it's a fairly obvious step once you get something that can operate iteratively and largely independent, there were a ton of people trying to get LLMs to loop on their own even before deepseek r1. And I remember talking about goal directed behavior (which what people are calling "agents" now don't seem to properly have) and autonomous operation decades ago in the intelligent agent course a…

> It's not very novel, though (...)

I completely disagree. The rollout of agentic tools, and even support for agent mode in IDEs, is the whole value proposition of AI code assistant services.

Otherwise you'd just have a glorified search engine in a chat window.

> (...) it's a fairly obvious step once you get something that can operate iteratively and largely independent,

There's some confusion in your reply. ReAct loops is exactly what this "operate iteratively and largely independently" represents.

Re: How LLMs work

#283

Post by lobocinza »

otabdeveloper4 wrote: post #169
jmalicki wrote: post #151

Earlier quoted context omitted.

Performance enhancements are huge though. If you can make the existing model faster, you can then save your inference budget to then make your model bigger, which then makes it smarter. A lot of how smart the models can be comes down to budget. If you can make your existing thing cheaper, you can instead make it bigger for the same price.

> to then make your model bigger, which then makes it smarter There's diminishing returns and at some point making a model bigger makes it dumber.

Maybe due to lack of data and dimensions other than words.

Re: How LLMs work

#284

Post by janalsncm »

wj wrote: post #235
janalsncm wrote: post #225

Earlier quoted context omitted.

The point of writing is to convey an idea to another person or yourself at a future date. Authenticity has nothing to do with it. I frankly do not care about the “authentic voice” of the author of a random blog. I want to know if they have any interesting ideas.

I think because so much of an idea is shaped by the language used to convey it, it may be hard to separate the person from the LLM. I think gp may want to know if a has an interesting idea rather than .

In that case, you can’t achieve that by writing things out by hand. A person could (and many do) use an LLM to generate an idea, then write it in their own voice. Or, they could use an LLM to write out an idea they came up with themselves.

In other words, since the idea generation component can completely independent from the writing component, what you’re asking is not possible in practice.

Re: How LLMs work

#285

Post by jmalicki »

Chu4eeno wrote: post #278
stevenhuang wrote: post #266

Earlier quoted context omitted.

You do realize modern neuroscience considers the human brain as "just" a probability engine and that intelligence may well be the ability for an organism to predict well. > doesn’t have a semantic understanding of what it is doing I hope you realize this is an area of open, active research.

Didn't neuroscience some big scandals about bad statistics and overstating their findings (in addition to normal issues like replication)? Look up at least the "dead salmon study" (hint: it's related to fMRI, and you can probably guess its conclusions from its nickname). The "Voodoo Correlations" and "Cluster Failure" papers are also a bit eye-opening. In general we (humans) need to be humble about the limitations of…

> In general we (humans) need to be humble about the limitations of our knowledge about how we function, it's an insanely complicated problem.

We do.

Which is why we shouldn't be assuming we're more than just probability engines, or be assuming we have more consciousness than a neural network.

Re: How LLMs work

#286

Post by layla5alive »

podocarp wrote: post #276
layla5alive wrote: post #249

Earlier quoted context omitted.

It's a literal truth that predicting the next token one at a time does not preclude intelligence on the other side of the decode function. Deal with it.

To be honest, if this is intelligence, then it's really boring. We can't even simulate the brain of a 1000 neuron nematode. You're telling me we can't even run nematode.exe but somehow we have already created intelligence?

Comparing either the Wright Brothers first successful flight (1903) or even an F22 Raptor (released in 2005) to a Sparrow:

"If this is flight, it's really boring. We can't even build a mechanical sparrow that can lay eggs and catch flies. You're telling me we can't run sparrow.exe but we've created flight?"

We didn't build something that flies by flapping its wings until 2010. We'd been building functional airplanes for more than 100 years before we were able to build something that worked in a (more primitive, but) similar way to how a sparrow flies.

I'm sorry that modern machine intelligence is so boring to you.

It isn't boring to me, I'm fascinated both by the ways I'm still far more capable than trillion parameter LLMs, and also by the ways they are already far more capable than I am.

FWIW, while I am not bored by nascent machine intelligence, I am bored by predictable human reactions to it: greed, exploitation, hubris, etc.

Re: How LLMs work

#287

Post by Borealid »

hackinthebochs wrote: post #277
Borealid wrote: post #273

Earlier quoted context omitted.

No, not nonsense. Both are a lookup table whose key is the entire context window and whose value is a probability distribution for what the next token should be. You can say the choice of probability distribution in the value is "leveraging the internal structure of the context" or not, but the same tokens in two different orders are two different lookup keys and saying it's impossible to achieve some result with a M…

That paper doesn't prove the equivalence of Transformers and Markov chains, it uses Markov chains as a theoretical model to understand the behavior of Transforms. The expressivity of the model matters, and Transformers just are more expressive than Markov chains. >but the same tokens in two different orders are two different lookup keys This is necessarily true for Markov chains and not necessarily true for Transform…

The paper presents a constructive transformation from any finite-input (finite vocab, bounded length) transformer to an equivalent Markov chain.

Do you have some concrete example of a transformer that cannot be represented as a mapping from inputs to probability distribution of outputs?

I say they're equivalent because it is possible to losslessly convert one to the other by wasting massive amounts of disk space and time.

As a second example proving the point, imagine you sampled a transformer's output for a certain context 85 trillion times, and put the output token frequencies in a table. Repeat for all possible inputs (of which there are a finite number). Then you built literally a hash map looking up the context and spitting out the distribution. That certainly is NOT a transformer any more (it's a hash map!!!), but the output approaches indistinguishability as the sample count increases - if the transformer is reasoning, so is the hash map built from it.

I'm not talking hot air here, they really are provably equivalent because a 1:1, onto mapping exists.

For the record, "X is more expressive than Y" means "there exists at least one thing that Y cannot represent and X can". Nothing to do with size or time.

Re: How LLMs work

#289

Post by hackinthebochs »

Borealid wrote: post #287
hackinthebochs wrote: post #277

Earlier quoted context omitted.

That paper doesn't prove the equivalence of Transformers and Markov chains, it uses Markov chains as a theoretical model to understand the behavior of Transforms. The expressivity of the model matters, and Transformers just are more expressive than Markov chains. >but the same tokens in two different orders are two different lookup keys This is necessarily true for Markov chains and not necessarily true for Transform…

The paper presents a constructive transformation from any finite-input (finite vocab, bounded length) transformer to an equivalent Markov chain. Do you have some concrete example of a transformer that cannot be represented as a mapping from inputs to probability distribution of outputs? I say they're equivalent because it is possible to losslessly convert one to the other by wasting massive amounts of disk space and…

>I say they're equivalent because it is possible to losslessly convert one to the other by wasting massive amounts of disk space and time.

There is a classical algorithm for every quantum algorithm if you're willing to waste a massive amount of space and time. There is a finite-state automata that can recognize any string some Turing machine can recognize. Yet we recognize these as distinct classes of computation. Mathematicians can get away with ignoring the tractability of finding an object with such and such properties. The rest of us can't.

Sure, there is a formal equivalence between LLMs and Markov chains, and this formal equivalence is useful for analysis. But this equivalence is not a constraint on the nature of the computations LLMs are doing. The formal equivalence does not mean that LLMs are "just predicting the next token". A probability distribution is a formal characterization of the statistical relationships between inputs and outputs. But this formalization does not undermine potentially further structure underlying the probability distribution (e.g. a deterministic mapping from inputs to outputs).

>if the transformer is reasoning, so is the hash map built from it.

Definitely not. "Formal" reasoning is making deductions based on the "form" or shape of some statement. In other words, transitioning from some token sequence to another sequence in virtue of the semantic structure of the token sequence (as opposed to its semantic content). Thus a necessary condition for reasoning is the ability to inspect the structure of the input rather than see it as a formless blob. Transformers can plausibly do this; lookup tables, Markov chains, etc necessarily cannot.

>For the record, "X is more expressive than Y" means "there exists at least one thing that Y cannot represent and X can".

Maybe expressive is the wrong word. But when a model has to wait for someone else to do the work then copy the answer, I call bullshit on it being (computationally) equivalent.

Re: How LLMs work

#290

Post by Borealid »

hackinthebochs wrote: post #289
Borealid wrote: post #287

Earlier quoted context omitted.

The paper presents a constructive transformation from any finite-input (finite vocab, bounded length) transformer to an equivalent Markov chain. Do you have some concrete example of a transformer that cannot be represented as a mapping from inputs to probability distribution of outputs? I say they're equivalent because it is possible to losslessly convert one to the other by wasting massive amounts of disk space and…

>I say they're equivalent because it is possible to losslessly convert one to the other by wasting massive amounts of disk space and time. There is a classical algorithm for every quantum algorithm if you're willing to waste a massive amount of space and time. There is a finite-state automata that can recognize any string some Turing machine can recognize. Yet we recognize these as distinct classes of computation. Ma…

Just to make sure I've understood you... Are you arguing that with a set of identically-behaving black boxes, one could be "reasoning" and one could be "not reasoning", and a person would need to look inside the boxes at how they function to decide?

Remember, if the mapping from input to output is identical, there exists no test operating on the machines' output that can differentiate them. You can't tell from "conversing with" a machine whether it is or is not doing what you say around "inspecting" the input.

Post reply on HN