emtel

I don’t know why you’re getting downvoted. One of my coworkers who was constantly touting Chinese models a few months ago now only uses opus 5.5. It is a huge difference.

Unless new data has come out, we don’t know if they are losing money on subscriptions. We know that there is a big difference between subscription and API costs per token, but that alone doesn’t mean much. Subscriptions may be underutilized, or margins on API tokens may be very high. Both are probably true.

This debate conflates two senses of the word deterministic:

1. Repeatable: the same inputs always lead to the same outputs 2. Algorithmic: the answer is arrived at by a set of steps that can clearly be shown to produce output with some desired property.

(2 is not a textbook definition of deterministic that I’ve ever heard, but I think it’s what people mean here)

LLMs don’t usually satisfy definition 1. They could, in principle, but it wouldn’t matter if they did. It wouldn’t tell you anything about whether the outputs are trustworthy.

What people seem to want is 2, but 2 is, I think, a contradiction. There cannot be an algorithmic process that can turns a natural language prompt into a working program, because natural language is too ambiguous to specify an output.

Nevertheless I agree with the main point of the article. Non-algorithmic processes (like intelligence) can produce algorithmic processes (like software). After all that’s what human programmers do!

It means that NNs don't (as previously thought) smoosh unrelated concepts together into the same neurons, which would have made interpretability very difficult.

For instance, if you had a single neuron that fires for "puppies" and "murder", it would be much more difficult to tell if the model is thinking something nice or something not so nice.

I’m not entirely unsympathetic to this view, but man. AI is here. It’s not going away. You may as well lament the fact that people use cars instead of horses.

The better use of your time is to start thinking about what new opportunities have opened up because of this. With AI coding your only limit is your imagination. Don’t waste it by sitting around and whining.

No evidence other than the fact that this has been happening steadily in all areas for many years?

You might have a point if the goal was to have LLMs that spit out a correct proof without chain of thought or tool use. LLMs + agent harnesses are more than capable of self verification and course correction.

There’s a small cottage industry of people who now have careers built on telling people that AI is a grift. Bender, Ed Zitron, Gary Marcus, etc. And there is a large audience who wants to hear this message.

What is troubling to me is I can’t think of a single prominent AI skeptic who has changed their mind. Are there any?

Are you serious? Lots of people disagree with what you are saying.

Today’s current problems were all hypothetical several years ago. At that time people claimed that the “real pressing problems” were misinformation and DEI issues. If we pretend that hypothetical problems can be safely ignored because there’s “no evidence” that they are real, we will keep getting surprised.

As far as I can tell, neither of the frontier US labs have referred to distillation as "stealing", but someone please provide a link if I'm wrong.

They do claim that it violates their ToS, which we can assume is simply correct, since they get to put whatever they want in their ToS.

Given all that, I don't know what the fuss is. Are they supposed to not use the advances that were openly published by Chinese labs? The entire industry is built on a discovery made at Google, which was published openly. Should Chinese labs therefore not use transformers? Should US labs not try to prevent distillation of their models?

You’re technically correct, but the vast majority of software has never been built to the kinds of standards you are describing. LLMs are not displacing that kind of work!

There is no limited supply of batteries. Production is growing and prices are declining rapidly.

Do you think people should continue developing AI up until the point that there is evidence that AI is facilitating biological weapons development?

I think we should stop before then. But that necessarily means that there will not be evidence at the time that we stop.

I have dabbled with Verus in rust. I found that if you leave the agent to its own devices it will produce a spec that just restates the existing implementation. It took a lot of work to arrive at specs that were plausibly accurate and complete statements of what the code was supposed to do.

But once you got there, Claude had no trouble proving the code satisfied the spec, and I uncovered a few interesting bugs this way.

I would say that it is not even wrong. You say it’s a next token predictor. I say it isn’t. What observable behavior of the system can settle our dispute?

I fail to see how any possible output could cause either of us to change our mind.

Can anyone explain why on earth you would counter bore a pipe at all, let alone after installation? I read the article but couldn’t understand the explanation. It seems like a totally ludicrous way to make a pipe.

> The model did exactly what it was told, albeit in an unintended, emergent strategy

Yes, that is the problem!

“Rare books” usually refers to rare editions of books. Any books out there where there are only a few extent copies of the text itself, are probably not of very much interest or social value, since almost no one is able to read them, by definition.

If you think there is priceless knowledge locked up in books so rare that it is on the verge of being lost forever, then AI labs are not really the problem!

In my experience fable can absolutely find very subtle bugs just by reading code and thinking. Last example I saw was a very subtle race condition that neither I not the author noticed. It wasn’t a security issue, but it could have been.

> Compiler performance is dominated by algorithms

But you can always use the best algorithm no matter what your implementation language is, so it still makes sense to prefer a language that makes it easy to write fast code.

> should I walk or drive to the car wash

there is an entire genre of riddles based on this kind of misdirection that works on humans, such as "A plane crashed on the border or US and Canada. Where do they bury the survivors?"

As with humans, comprehension is not a binary property of the agent - it is a quality that can be present in some situations and absent in others. LLMs may emit correct outputs sometimes because they do comprehend the input, and emit incorrect outputs in other cases when they do not comprehend the input.

In order to show that LLMs can't comprehend, we'd have to show that there are no (or at least very few) situations in which they exhibit comprehension, not show that there are some situations in which they don't.

Even if pre-training was the only training step, it still wouldn't necessarily follow that the only thing the model is doing is stitching words together probabilistically, unless you expand the definition of "probabilistically" to the point that it becomes meaningless. This kind of thinking assumes that design of the training process and the "design" of the artifact that training produces must be similar.

I’m guessing you don’t have kids. My membership pays for itself just for the savings on string cheese alone. Same for eggs.

audiophiles (https://forums.stevehoffman.tv/threads/turntables-with-pace....) also claim that turntables can be rated on "timing, rhythm, and pace" in which supposedly the timing of the music can be affected by the turntable's mass and other properties.

How this would occur without also producing grossly audible pitch distortion never seems to be discussed.

What a strange paragraph! Why on earth should we assume that a body is a prerequisite for having emotions?

How is drugging an animal's food a "brute force tactic"? What would qualify as a non-brute-force tactic in your mind?

The "Einstein can't get a cat into a carrier even though he is smarter" is just a hilariously bad argument. All cat owners can get their cats into a carrier! And most cats don't want to get in, because they hate the vet! And it's almost entirely because the humans are smarter!

You can even get a literal tiger into a carrier, even though it can kill you easily. You just drug its food and wait till it passes out. This is because you are smarter than it, and know that tranquilizers exist and how to obtain them, which is a strategy that cats of any size are not even able to conceive of, and probably can't understand what happened after it's been done to them.

A lot of posts urging people to get colonoscopies. But general anesthesia also carries risks. How should one weigh the risk vs benefit?

> The goal is not to force individuals to not replace their phones, but rather to provide that as an option at all, for those who want it.

But my point is that you need to recognize that in so doing, you are taking away the option of having other things, such as waterproofing, larger batteries, smaller/lighter phones, etc. There is no free lunch.

That's an odd reply since by that argument they also flocked to a phone with no replaceable battery, which was pretty standard in the 2000s.

But you could be right. I guess this will be an experiment to watch: If EU consumers show a strong preference for replaceable batteries once they become more widely available, we can expect manufacturers to start offering it in other markets as well.