But we know exactly what they’re doing conceptually, and individually, and in aggregate.
Can you define and give examples of what you mean at each level here? Maybe we're just not understanding each other and mean the same thing.
Read your own sources from your previous post, that’s what they’re telling you.
The Anthropic one is saying they think they have a way to figure it out, but it hasn't been tested on large models. This is their last paragraph:
Our next challenge is to scale this approach up from the small model we demonstrate success on to frontier models which are many times larger and substantially more complicated. For the first time, we feel that the next primary obstacle to interpreting large language models is engineering rather than science.
They are literally only able to do this on a small one layer transformer model. GPT 3 has 96 layers and 175 billion parameters.
Also, in their linked paper:
A key challenge to our agenda of reverse engineering neural networks is the curse of dimensionality: as we study ever-larger models, the volume of the latent space representing the model's internal state that we need to interpret grows exponentially. We do not currently see a way to understand, search or enumerate such a space unless it can be decomposed into independent components, each of which we can understand on its own.
Under the Future Work heading:
Scaling the application of sparse autoencoders to frontier models strikes us as one of the most important questions going forward. We're quite hopeful that these or similar methods will work – Cunningham et al.'s work [17] seems to suggest this approach can work on somewhat larger models, and we have preliminary results that point in the same direction. However, there are significant computational challenges to be overcome.
How are you getting from that that this is a solved problem?
Concepts are indeed abstract but LLMs have no concepts in them, simply vectors. The vectors do not represent concepts in anything close to the same way that your thoughts do. They are not 1:1 with objects, they are not a “thought,” and anyway there is nothing to “think” them. They are literally only word weights, transformed to text at the end of the generation process.
Again, you aren't making sense here. Word/sentence vectors are literally a way to represent the concept of those words/sentences. That's what they were built for. That's how they are described. Let's take a step back to try to understand each other.
Are you trying to say that only human minds can understand concepts? I don't buy the human brains are magic bit, and neither does our current understanding of physics. Are you assuming I'm saying that LLMs are sentient, conscious, have thoughts or similar? I'm not. Jury's out on the thought thing, but I certainly don't believe the other two things. There's no magic with them, same with human brains. We just don't fully understand what happens inside either. Anthropic in the work I quoted is making good progress at that, and I think they may be pretty close, but in terms of LLMs (and not Small LMs), they are still a black box. We know the math behind them, the software, etc. We have some theories. We still do not understand. If you can prove otherwise, please provide me with a source. Stuff is happening really fast in AI, and maybe I blinked and missed something.
I think you're maybe having a hard time with using numbers to represent concepts. While a lot less abstract, we do this all the time in geometry. ((0, 0), (10, 0), (10, 10), (0, 10), (0, 0)) What's that? It's a square. Word vectors work differently but have the same outcome (albeit in a more abstract way).
the vectors do not represent single words, but tokens
I was talking word vectors where the vectors DO represent words. It's in the name. LLMs don't specifically use word vectors, but the embeddings they do use work similarly.
Tokens do not represent the meaning of a word/partial word/phrase, just the statistical use of that word given the data the word was found in.
You are correct tokens don't represent the meaning of a word. However, tokens are scalars. You are conflating tokens and embeddings / word vectors here. Tokens are used to simplify converting a string into a format a neural network can understand (a vector). If we used each ascii character in the input/output string as a vector input to the network, we'd have to have a lot more parameters than if we combine the characters in some way (i.e. tokens). As you said, they can be a word or a part of a word. There's no statistics embedded with the tokens (there are some methods of using statistics to choose what tokens to use, but that's decided before even training the model and can not ever change [with our current approach]). You can read here for more information on tokens. Or you can play around with the gpt3 tokenizer.
Your concept of a chair is an abstract thought representation of a chair. An LLM has vectors that combine or decompose in some way to turn into the word “chair,” but are not a concept of a chair or an abstract representation of a chair. It is simply vectors and weights, unrelated to anything that actually exists.
If you know Python, you should grab nltk and experiment with gensim, their word vectors.
model.most_similar(positive=['woman','king'], negative=['man'], topn = 1) [('queen', 0.71181...)]
king + woman - man = queen
Seems like an abstract representation of those things as concepts using math. For the record, word vectors are actually pretty understandable/understood by people because you can visualize them easily. When you do, you find similar concepts clustered together (this is how vector search works except with text embeddings). Anyway, it just really seems like linking numbers to concepts is not clicking with you, or you somehow think it's not possible. Reading up on computational linguistics might help.
That is obviously totally different in kind to human thought and abstract concepts. It is just not that, and not even remotely similar.
Yes, neural networks (although initially built thinking they were a computer version of a neuron), are a lot different from how actual brains work as we've learned in however many decades since they were invented. If you're saying that intelligence and understanding is limited to the human mind, then please point to some non-religious literature that backs up your assertion.
You say you are familiar with neural networks and AI but these are really basic underpinnings of those concepts that you are misunderstanding. Maybe you need to do more research here before asserting your experience?
I'm pretty confident in my understanding, though I'm always open to new ideas that are backed with peer reviewed research. I'm not going to get into a dick waving contest here, so I guess we'll have to agree to disagree.
As a side note, going back to your definition of intelligence. That was for psychology. I'll note that the Wikipedia page for Intelligence has this to say:
The definition of intelligence is controversial, varying in what its abilities are and whether or not it is quantifiable.
And so I'll reiterate that we don't have a good definition of intelligence.
Wait a sec. I think we're saying the same thing here. I guess depending on what you mean by how it operates and functions. I've said multiple times we understand the math and the code. We understand how values propagate through it because again, that's all the math and code people wrote. What we don't understand is how it uses that math and code to actually do thinks that seem intelligent (putting aside the point of whether it is or is not intelligent). If that's what you're arguing then great, we're on the same page!
Well, I don't have the equipment to look at electrons either (I don't think that tech exists), but I can take a logic probe and get some information that I could probably understand, or someone who designs CPUs could look at the gates and whatever and tell you what they did and how they relate to whatever higher level operations. You're bringing up something completely different here. Computers are not a black box at all. LLMs are-- you just said that yourself.
I'm not anthropomorphisizing them. What are you talking about? I keep saying they don't work like human brains. I just said I don't think they're sentient or conscious. I said they don't have agency.
How do you know what it's not if we can't define what it is?
Jury's still out on whether human brains are complex statistical models. I mean (from here)...
I don't make any claim to understanding neuroscience, and I don't think that article is saying for sure we know that.
Anyway, in-context learning is a thing for LLMs. Maybe one day we'll figure out how to have them adjust their weights after training, but that's not happening now (well people are experimenting with it).
New research is showing they do have semantic understanding.
They don't by themselves have self-awareness, but a software framework built up around them can generally do that to some extent.
They do understand emotions and ethics. Someone built a fun GPTrolley web site a while ago. I think it died pretty quickly because it was too expensive for them, but it had GPT 3(?) answering Trolley Problem questions. It did (in my memory of it) like to save any "AGI" on one track over humans, which was amusing. They don't have emotions, no. Does something have to have emotions to be intelligent?
And no, I've said all along they aren't conscious, so no qualia. Again, is that required for intelligence?
No. A square to GPTs is not just a token. It's associated with some meaning. I'm not going to re-hash embedding and word vectors and whatever since I feel like I've explained that to death.
I'm literally not. "Intelligence is limited to the human mind" is not a negative.
I feel like I've laid out my argument for that mostly through the Microsoft and Max Tegmark papers. Are you saying intelligence is only the domain of biological life?
Here's a question-- are you conflating "intelligence" with "general intelligence" like AGI? I find a lot of people think "AI" means "AGI." It doesn't help that some people do say those things interchangeably. I was just reading a recent argument between Yann LeCun and Yoshua Bengio and they were both totally doing that. Anyway, I don't at all believe GPT4 is AGI or that LLMs could even be AGI.
Looks like a great paper-- I hadn't seen it yet. I know how LLMs are constructed (generally-- while I could go and write some code for a multi-layer neural network with back propagation without looking anything up, I couldn't do that for an LLM without looking at a diagram of the layers or whatnot).