Exactly! Like individualized and infinitely patient always available personal tutors. Oh, wait…

You got a source for this?

You seem reasonable and even-tempered, so I’d like to give you a considered response so I can hear your thoughts. But first I wanna ask:

Do you object to the technology as a whole, or only to the use of the moniker AI to described LLMs?

What is your definition of intelligence?

I love the appropriate application of Kant!

But I think the point about bad faith misses the mark. For one to conduct themselves or argue for a position in bad faith (even against only themselves), they must first hold a true, hidden belief. I don’t think that’s typically the case here.

You could argue that there is a hidden belief and that it is something like “AI is bad” or “I don’t like AI.” However, the people arguing this position are often quick to point out themselves that AI has legitimate value, even if they then go on to narrowly define the character and application of that value.

I would argue instead that what is actually being hidden here isn’t a belief, but is instead a sentiment something like, “I’m scared.” Which, while being completely understandable, is an emotive state or pseudo-belief, not a truly-held but hidden belief. Because of that, they (in the general sense) aren’t calculating hypocrites; they’re are just panicking.

Although, I prefer the admittedly archaic flatus vocis for describing this particular brand of cognitive dissonance because it translates basically to “mouth farts,” and that makes me chuckle.

1. With greedy decoding (temperature set to 0) LLM output is deterministic. The same input will always result in the same output — every time. This is a fact.
2. Just because something isn’t amenable to calculation does not mean it is nondeterministic, for example the Three Body Problem. This is a fact.
3. Humans are stupid unpredictable base animals that hold almost no understanding of the world around them or their place within it, while at the same time their heard-level behaviors are almost, if not entirely, predictable. This is a cynical joke.

so there is a field of ai. … even partial steps have wide reaching applications and philosophical implications. it is the absolute coolest shit, and gets exactly 0 funding these days.

the thing all the data centers RAM water electricity and economy are backing right now is not ai. it’s what happens when a magic trick and a ponzi scheme fuck on a pile of cocaine in a Pharaoh’s tomb.

These two positions seem to be mutually exclusive, and your response is, on the whole, incoherent. Could you please explain what it is you are trying to communicate?

This is almost exactly the opposite of how LLMs work.

LLMs running under greedy decoding are deterministic finite state machines. A lot of work has gone into making their output nondeterministic because probabilistic next-token generation is more useful.

The more accurate characterization would be that LLMs are context-dependent deterministic state machines with adjustable nondeterministic sampling algorithms bolted onto the decoder.

Also, just as a side note:

They have no determinism, no concept of context and their “reaction” are pure statistics.

This almost perfectly describes human behavior.

They are statistical garbage generators with a high word prediction accuracy that is enough to fool humans into falsely believing they have understanding of context.

That very precisely describes a lot of people I know.

Could you inform me of what is really going on here?

I honestly don’t know what point you are trying to make, but since you are clearly confused by my comment, let me try to further explain.

My claim: The observation that you can’t have space exploration without ballistic missiles is correct. This analogy as applied to AI (specifically LLMs) is also correct, but falls short of perfect parity.

The gap — which is what I tried to address in my previous comment — is that for the aerospace example, space exploration vehicles are like ballistic missiles, but for LLMs those trained to do medical research or those trained to be a chat bot are often not just similar, but the exact same.

That was the point. An LLM built for research vs one built for chat are not just similar, like an ICBM vs a manned space vehicle, but are literally exactly the same object with no difference whatsoever.

This was my attempt to support the person I responded to against the critically uninformed, “Why don’t we just have the one (medical research AI) without the other (chat bot)?” crowd. With the obvious answer being, “Because they are the same thing.”

Q: So how’s it different? A: It’s not. That was my point.

Yeah, fair enough.

I’m sorry, but it’s not my responsibility to educate you about copyright law, even though I have done what I could. I wish you the best. Good day.

When i write a sentence then do i have a copyright upon it which i can defend in law?

Umm… Yes. Yes, you do. That’s how copyright works…

The current models are all trained with data which at least some larger creator (book writer or whatever) does not want a model to be trained on.

Is this not antithetical to your stance? Or are you saying that only those entities large enough to defend themselves against the likes of OpenAI and Anthropic are deserving of being paid for use of their works?

The current models are under the current law unsustainable.

I completely agree with you here, but something like a class action lawsuit is not a viable solution.

You understand the scale right? This is a combinatorial problem: the plaintiffs in this case would be every individual and company that has ever existed, and the defendants would be not only the AI companies, but also anyone who has ever used or published AI generated content. That’s a wishful thought, not a realistic solution.

Holy shit! What an excellent idea.

I always advocate for putting AI models under UNESCO stewardship

As now will I.

Copyright isn’t something you “get.” It’s something that comes with the production, no extra steps required.

You’ve accidentally come upon the other reason why your example doesn’t hold in the training data case. How do you suss out who actually holds the copyright on that data.

Take my shitpost for example. Whonis the copyright holder? Me the author? Reddit the platform I published on? What if the source was an Internet Archive snapshot?

Do you earnestly believe this could work in practice?

And we are talking about meaningful data not your shitpost, like licensed code or a book you wrote and anthropic pirated.

Ah ha! You see, that’s the problem. My shitpost is meaningful data. This is something people seem to have a really hard time understanding. Scaling laws and the size of the search space in AI training means that the datum is not valuable. It’s only the data (in conglomerate) that is value for training, and my shitpost would be a part of that.

Once you understand this, you should see what I meant when I said it’s not feasible. How do you determine and then fairly distribute the payout to everyone who has ever written or recorded anything in digital format? This is not a simple supply vs demand equation, like your bakery example assumes.

Also I have no idea what Netflix has to do with this discussion.

How do you suggest an individual should go about this?

Do I go down to my local courthouse and file a suit against Anthropic so I can ask the judge to subpoena them to determine if my Reddit shitpost from 2012 is in their training set?

How is what people do any better or different than the Really Good Guessing Machine?

I really don’t understand this perspective, and it’s my biggest gripe with Chollet and his ilk as well. What, in your opinion, makes neuro-symbolic AI inherently superior to (the currently better performing) statistical methods?

I don’t think it’s really possible to do this.

I mean speaking just logistically, how would you determine who gets how much? If it’s based on quality, how do you determine what is of high quality? Otherwise, does everyone who has every written anything that finds its way into a training dataset get like a tenth of a cent?

Alternative proposal: all AI models must be released as open weights, available to everyone.

AI, through its use of training data, has benefited from the sum total of human knowledge, therefore the results of this training should be given back for the use of all humanity.

That seems fair to me.