this post was submitted on 30 May 2024
1099 points (98.8% liked)
Funny
6773 readers
680 users here now
General rules:
- Be kind.
- All posts must make an attempt to be funny.
- Obey the general sh.itjust.works instance rules.
- No politics or political figures. There are plenty of other politics communities to choose from.
- Don't post anything grotesque or potentially illegal. Examples include pornography, gore, animal cruelty, inappropriate jokes involving kids, etc.
Exceptions may be made at the discretion of the mods.
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Wow, that's actually quite impressive.
I'm sure eventually someone will make a bot called something like ai-explains-the-joke that does this automatically.
Expl-AI-n Bot will break down whatever joke you feed it.
Expl-AI-n itself is a pun. With the letters AI in the word explain capitalized, readers can infer that artificial intelligence is being used to explain jokes.
I wonder how much was scraped from knowyourmeme.com
I mean it still parsed the specific text in the meme and formulated a coherent explanation of this specific meme, not just the meme format
Or it matched the text with an existing explanation upon which it was indexed.
Lmao you think it found a specific explanation for this specific variation of this meme?
For each phrase, yes.
That's not how GPTs work
That's literally how they work
Man the models can't store verbatim its training data, the amount of data is turned into a model that is hundreds or thousands of times smaller than the original source data. If it was capable of simply recovering everything that it was trained on this would be some magical compression algorithm and that by itself would be extremely impressive.
Congratulations on discovering compression
Oh ok, you want to claim this is compressing the entirety of the internet in a model that isn't even 1 terabyte of data and be unimpressed that is something.
But it isn't compression. It is a mathematical fact that neural networks are universal function approximators, this is undisputed, and analytic functions are continuous so to be an analytical function approximator it must be able to fill in the gaps between discrete data points by itself, which necessarily means spiting out data outside of the input distribution, data it has not seen.
TBF, compression is related to ML. Hence, the Hutter Prize. Thinking of LLMs as lossy compression algorithms is a decent analogy.
It is a partial analogy, it takes into consideration the outputs which are related to some specific training data and disconsiders the outputs which cannot be directly related to any specific training data.
For example, make up a new meme template and a new joke on the spot, it couldn't have seen it before if you make sure your joke and template are new. If the AI can explain it then compression is a horrendous analogy.
Lossy compression explains outputs being similar but not identical when trying to recover the original data, it doesn't explain brand new content that makes sense standalone. Imagine a lossy audio compression resulting in a brand new song midway through playback, or a lossy image compression resulting in a brand new coherent image being overlayed onto some pixels of the original image. That is not what happens, lossy audio compression results in noise, lossy image compression results in noise, not in coherent unheard songs and unseen images.
Not sure why you feel the need to put words in my mouth. It wasn't trained on "the entirety of the Internet," but rather less than a terabyte of it. So yeah, that would probably take up less than a terabyte.
Then why did I just make this meme up right now and chat gpt can explain it?
https://i.ibb.co/NYHRnTY/Screenshot-20240531-072008-Chat-GPT.jpg
Arguing over this is just dumb, you can yourself take any picture you want at this very moment or come up with a brand new meme template on the spot and upload it to ChatGPT to see you are wrong, it is free btw.
They do not store anything verbatim; They instead store the directions in which various words and related concepts relate to one another in some gigantic multidimensional space.
I highly suggest you go learn what they actually do before you continue talking out of your ass about them
If you trained a GPT on a single phrase, all you'd get out of it would be the single phrase.
The mechanism of storage doesn't need to be just the verbatim source material, which is not even close to what I said.
You said it matches text to its training data, which it does not do.
Your single-phrase statement only works for very short, non-repetitive phrases. As soon as your phrase repeats a token more than a few times, the statistics for the tokens change and could result in nonsensical output that repeats through subsections of the training data.
And even then for that single non-repetitive phrases, the reason you would get that single phrase back is not because it would be "matching on" the phrase. It is because the token weights would effectively encode that the statistical likelihood of the "next token" in the generated output is 100% for a given token when the evaluated token precedes it in the training phrase. Or in other words: Your training data being a single phrase maniplates the statistics so that the most likely output is that single phrase.
However, that is a far cry from simple "matching" against the training data. Which is what you said it does.
If it doesn't use its training data, what's the training data for?
Analysis. It uses it, but not by "matching it". The training data is not included in the final model. No GPT can access its training data at runtime.
Training analyzes the contents of the training data and creates a statistical model representing the likelihoods of various tokens based on a complex series of mathematical transformations that encode various attributes of the tokens making up the training data.
3Blue1Brown has a great series on the actual math behind it, I would highly recommend educating yourself on what GPTs actually do. It's way more interesting than simple matching.
God forbid I use simpler language to describe what it does.
Simpler language is fine when it's accurate.
Your simplification is inaccurate and could mislead people into thinking GPTs are just advanced regex matching engines.
They are not. They are closer to autocorrect on steroids.
Autocorrect is fancy pattern matching. GPT is fancier pattern matching.
It's more accurate than "AI," since there's no actual reasoning happening.
I'm gonna stop responding to this asanine thread now before you continue to demean us both with your nonsense.
The majority of people right now are fairly out of touch with the actual capabilities of modern models.
There's a combination of the tech learning curve on the human side as well as an amplification of stories about the 0.5% most extreme failure conditions by a press core desperate to feature how shitty the technology they are terrified of taking their jobs is.
There's some wild stuff most people just haven't seen.
I can just as well say that the screenshot above is the top 0.5% pushed by people trying to sell the tech. I don't really have an opinion either way tbh, I'm just being cynical. But my own experience with those tools hasn't been impressive.
At a pretrained layer, the model is literally a combination of a normal distribution curve of capabilities.
It can autocomplete a flat earther as much as a Nobel physicist given sufficient context.
So it makes sense that even after the fine tuning efforts there'd be a distribution in people's experiences with the tools.
But just as the average person's output from Photoshop isn't going to be very impressive, if all you ever really see is bad Photoshops and average use, you might think it's a crappy tool.
There's a learning curve to the model usage, and even in just a year of research the difference between capabilities of the exact same model from then to now is drastically different, based only on learnings around better usage.
The problem is the base models are improving so quickly the best practices for the old generation of models goes out the window with the new. So even if there were classes available I wouldn't bother pointing you to them as you'd just be picking up info obsolete by the time the classes finished or shortly thereafter.
I'd just strongly caution against betting against the tech's continued capabilities and improvements if you don't want to be surprised and haven't taken the time to look into them operating at their best.
The OP post is pretty crap compared to the top 0.5% usage.
It really does depend on what you ask and how, I can get some really nice music recommendations from Chatgpt but it also cannot comprehend GURPS skill rules, it's actually funny how it manages to get it wrong a completely different way each time
At the risk of sounding like a tech bro who's desperately trying to secure funding: this truly does feel like a major leap in technology that is going to change the world.
Anytime I hear it dismissed as "basically auto-complete", I feel like it's being underestimated.
It's not just underestimation, it's outright misinformation.
There's so much research by this point over the past 18 months that there's an incredible amount going on beyond "it's just a Markov chain, bro."
It was never a Markov chain as that ignored the self-attention mechanism which violated the Markov property. It was just some people trying to explain it used a simplified description which went viral.
Its kind of funny because autocomplete on phones is definitely moving in the direction of using LLMs. Its like it wasn't true when people started saying it, but it will be literally true in a couple of years at most.