(Related text posted to Twitter; this version is edited and has a more advanced final section.)
Imagine yourself in a box, trying to predict the next word - assign as much probability mass to the next token as possible - for all the text on the Internet.
Koan: Is this a task whose difficulty caps out as human intelligence, or at the intelligence level of the smartest human who wrote any Internet text? What factors make that task easier, or harder? (If you don't have an answer, maybe take a minute to generate one, or alternatively, try to predict what I'll say next; if you do have an answer, take a moment to review it inside your mind, or maybe say the words out loud.)
Consider that somewhere on the internet is probably a list of thruples: <product of 2 prime numbers, first prime, second prime>.
GPT obviously isn't going to predict that successfully for significantly-sized primes, but it illustrates the basic point:
There is no law saying that a predictor only needs to be as intelligent as the generator, in order to predict the generator's next token.
Indeed, in general, you've got to be more intelligent to predict particular X, than to generate realistic X. GPTs are being trained to a much harder task than GANs.
Same spirit: <Hash, plaintext> pairs, which you can't predict without cracking the hash algorithm, but which you could far more easily generate typical instances of if you were trying to pass a GAN's discriminator about it (assuming a discriminator that had learned to compute hash functions).
Consider that some of the text on the Internet isn't humans casually chatting. It's the results section of a science paper. It's news stories that say what happened on a particular day, where maybe no human would be smart enough to predict the next thing that happened in the news story in advance of it happening.
As Ilya Sutskever compactly put it, to learn to predict text, is to learn to predict the causal processes of which the text is a shadow.
Lots of what's shadowed on the Internet has a *complicated* causal process generating it.
Consider that sometimes human beings, in the course of talking, make errors.
GPTs are not being trained to imitate human error. They're being trained to *predict* human error.
Consider the asymmetry between you, who makes an error, and an outside mind that knows you well enough and in enough detail to predict *which* errors you'll make.
If you then ask that predictor to become an actress and play the character of you, the actress will guess which errors you'll make, and play those errors. If the actress guesses correctly, it doesn't mean the actress is just as error-prone as you.
Consider that a lot of the text on the Internet isn't extemporaneous speech. It's text that people crafted over hours or days.
GPT-4 is being asked to predict it in 200 serial steps or however many layers it's got, just like if a human was extemporizing their immediate thoughts.
A human can write a rap battle in an hour. A GPT loss function would like the GPT to be intelligent enough to predict it on the fly.
Or maybe simplest:
Imagine somebody telling you to make up random words, and you say, "Morvelkainen bloombla ringa mongo."
Imagine a mind of a level - where, to be clear, I'm not saying GPTs are at this level yet -
Imagine a Mind of a level where it can hear you say 'morvelkainen blaambla ringa', and maybe also read your entire social media history, and then manage to assign 20% probability that your next utterance is 'mongo'.
The fact that this Mind could double as a really good actor playing your character, does not mean They are only exactly as smart as you.
When you're trying to be human-equivalent at writing text, you can just make up whatever output, and it's now a human output because you're human and you chose to output that.
GPT-4 is being asked to predict all that stuff you're making up. It doesn't get to make up whatever. It is being asked to model what you were thinking - the thoughts in your mind whose shadow is your text output - so as to assign as much probability as possible to your true next word.
Figuring out that your next utterance is 'mongo' is not mostly a question, I'd guess, of that mighty Mind being hammered into the shape of a thing that can simulate arbitrary humans, and then some less intelligent subprocess being responsible for adapting the shape of that Mind to be you exactly, after which it simulates you saying 'mongo'. Figuring out exactly who's talking, to that degree, is a hard inference problem which seems like noticeably harder mental work than the part where you just say 'mongo'.
When you predict how to chip a flint handaxe, you are not mostly a causal process that behaves like a flint handaxe, plus some computationally weaker thing that figures out which flint handaxe to be. It's not a problem that is best solved by "have the difficult ability to be like any particular flint handaxe, and then easily figure out which flint handaxe to be".
GPT-4 is still not as smart as a human in many ways, but it's naked mathematical truth that the task GPTs are being trained on is harder than being an actual human.
And since the task that GPTs are being trained on is different from and harder than the task of being a human, it would be surprising - even leaving aside all the ways that gradient descent differs from natural selection - if GPTs ended up thinking the way humans do, in order to solve that problem.
GPTs are not Imitators, nor Simulators, but Predictors.
I think the devil may be in the details here. Ask GPT to hash you a word (let alone guess which word was the origin of a hash), it'll just put together some hash-like string of tokens. It's got the right length and the right character set (namely, it's a hex number), but otherwise, it's nonsense.
This ties into the whole "does it understand" question because it's a very simple example of a simple prediction question with a very deep underlying complexity in which a GPT doesn't perform much better than an N-gram Markov chain. Because there is a lot of complexity to the task that isn't exemplified at all in the simple password-hash pair, and no matter how many password-hash pairs you get, inferring the hashing algorithm is hard. So you end up just mimicking hash-like sequences and calling it a day. You parrot, but you don't understand.
How much of the rest does GPT-4 understand, though? It seems to have a thorough enough world model to at least use words always in sensible and appropriate ways. It seems to understand their meaning, if in a purely relational sense: <adjective> is a property of <noun>, which can perform <verb> or <verb>, and so on. That's something where there's a lot more examples, and the underlying complexity isn't nearly as much as the hashing algorithm (after all, language is designed to make the meaning clear, hashing is designed to obfuscate).
There's a question of which information is implicitly contained in a text, and which just isn't. If you ask GPT-4 to write a new scientific paper it simply can't come up with new empirical results, no matter how smart it is, if they're not implicitly derivable from its training set (that is, they're not actually novel, just an overlooked implication of existing knowledge). So that's a task no text prediction can possibly be up to, no matter how smart. Hashing is theoretically included in the training set (there's descriptions of the algorithm in there!) so not being able to do it definitely qualifies as "GPT-4 is too stupid to do this". Reverse hashing is different because there's no known general solution to it, so being able to do it in a few shots would make it vastly superhuman, but I can imagine there existing eventually an AI that can do it with a good success rate.
I suppose you could also make a distinction here between model and simulacrum. The model is an unfathomable shoggoth who is trained at a superhuman task, and is much better at it than most humans. The simulacra, aka the fake personalities that emerge when the model is asked to predict a dialogue between set characters, may be a lot stupider than that because they're pretend humans talking; they don't have prediction abilities, they are the result of predictions. ChatGPT as we know it isn't a model, but a simulacrum. The model is the invisible puppet master moving it behind the scenes.
Base64 encoding is a substitution cipher. Large language models seem to be good at learning substitutions.