> The goal of character training is to make Claude begin to have more nuanced, richer traits like curiosity, open-mindedness, and thoughtfulness.
You can't make a transformer-based language model "have" curiosity.
Real curiosity is an innate trait in intelligent animals that promotes learning and exploration by increasing focus and stick-to-it-ness when the situation is interesting/unexplored (i.e. not well predicted - surprising).
An LLM could be trained to fake curiosity in same way that ELIZA did ("Tell me more about your fear of prison showers"), but fundamentally it deals in words and statistics, not facts and memories, so any poorly-predicted surprise signal will be surface level such as "that was an unusual word pattern" or "you haven't mentioned that before in the input context".
Even if fake curiosity ("interest") makes the model seem more conversationally engaging, maybe helps exploration by prompting the use to delve deeper into a subject, it's not going to help learning since LLMs are pre-trained and cannot learn by experience (except ephemerally within context).
This arguments is made every singe time a new LLM article gets posted and I'm not sure it's really adding anything to the conversation. Everyone understands language models are not human, there's no need to add a philosophical argument because you don't like the word choice.
"Characters" in any other context are also by definition not curious. They're not open-minded or thoughtful either. Characters in a book are not people, they don't have thoughts, they are whatever the author put on paper. Yet we still use these words to describe them, as if the characters' consciousness and motivations existed beyond the paper on which they're described. It becomes really hard to talk about these traits without using words like curiosity and thoughtfulness. No one thinks the characters in a fictional book are real people.
I'm not sure that "these are just characters, your expectations are too high" is totally valid here, especially as Anthropic are claiming their LLM progression is on the path to AGI.
Any true human-level (or even rat-level for that matter) AGI would need to have actual innate traits such as curiosity to drive lifelong exploration and learning. That seems a pretty minimal bar to meet to claim that something is AGI.
I suppose in context of an article about giving Claude a character (having it play a character) then we need to interpret "having a trait" as "playing a character having a trait", because it certainly is very different from actually having it.
> Any true human-level (or even rat-level for that matter) AGI would need to have actual innate traits such as curiosity to drive lifelong exploration and learning.
While perhaps not "true" intelligence (whatever that is), with LLMs this can be emulated with a prompt:
> You have an innate curiosity to drive lifelong exploration and learning. ...
If the behavior of the system is indistinguishable from "true" intelligence, the distinction becomes a philosophical one.
How can you emulate learning when LLMs have no way to update their weights? LLMs don't even know what they do and don't know (hence hallucinations), so how would they even realize that something was new in the first place?
I could maybe see chatbots storing everything you ever said in a vector database for future lookup, but that's just memory.
There's nothing stopping you as an LLM system designer from having it "go to sleep" once a day and use that downtime to consolidate the memory in the vector database into fine-tuning of the weights.
Some humans who are not trained that curiosity produces beneficial outcomes are often not all that curious. It is difficult to know what is truly innate vs. the result of conditioning. It is also difficult to know what is the result of an absence of opposing conditioning. If I was raised in a poor area with high crime and little economic opportunity, my conditioning would be quite different. How would that have changed my behavior? Could curiosity get me in more trouble than it's worth? Could I observe that and adjust my own behavior as a result?
We are all, to some extent, a product of our environment (training data). I wasn't raised in that type of area, but does that mean my own intellectual curiosity is more innate? Or does it mean it is less innate? I could argue that both ways.
> Some humans who are not trained that curiosity produces beneficial outcomes are often not all that curious.
Innate traits such as curiosity & boredom are things that we are born with, not learnt. The reason evolution has selected for these innate traits is because there is a benefit (encouraging exploration and learning which help survival), but you don't need to be aware of the benefits to experience and act on boredom or curiosity.
Innate behaviors can certainly be reinforced, or suppressed to some degree, by experience.
I think you _can_ make an LLM 'have' curiosity, for all practical intents and purposes.
I'm e.g. thinking of the 'have the LLM research a topic' task, such as the 'come up with suitable search terms, search the web, summarize the article, then think of potential next questions' cycle implemented by Perplexity, for example. I'm pretty sure the results would vary noticeably between an LLM that was trained to be 'curious' (i.e., follow more unusual trains of thought) versus not, and the differences would probably compound the more freedom you give the LLM, for example by giving it more iterations of the 'formulate questions, search, summarize' loop.
The problem is how can "follow more unusual trains of thought" apply to a language model ? Sure it can selectively attend to certain parts of the input and generate based on that, but what is the internal signal for "unusual" ? Any selective focus is also going to appear like groundhog day since the model's weights are fixed and what was "unusual" today will continue to be "unusual" even after it's been exposed to it for the 1000th time!
Thinking about this a bit, it might be a bit late actually to start to guide an LLM towards curiosity only at the fine-tuning stage, since this 'exploring unusual-trains-of-thoughts' is precisely what the LLM _isn't_ learning during training, where it sees (basically by definition) a ton of 'usual trains-of-thoughts'. Maybe you'd have to explicitly model 'surprise' during training, to get the LLM to try to fit precisely those examples better that don't really fit its already learned model (which would require the network to reserve some capacity for creativity/curiosity, which it otherwise might not do, because it's not necessary to model _most_ of what it sees). But then you enter the territory of 'if you open your mind too much, your brain might fall out', and could end up accidentally training QAnonGPT, and that you definitely don't want...
So maybe this way of 'hoping the LLM builds up enough creative intelligence during training, which can then be guided during fine-tuning' is the best we can do at the moment.
Curiosity is inherently proactive. An LLM, fundamentally, is a file. It's a FILE. It just sits there until you ask for output. Saying anything more about that process is just anthropomorphizing.
No one ever accused the Google index of having "curiosity", but the idea is basically the same - you give it a query and it gives you back a response. Then it just sits idle until the next query.
> Saying anything more about that process is just anthropomorphizing.
Take "this LLM is more curious" as a shorthand for "the output generated by this LLM mimics the kind of behaviour we would describe in a human as being curious".
> It just sits there until you ask for output.
That is indeed a property of the current interfaces. But this can be very easily changed. If we choose to we can just pipe in the clock, and then we can train the model to write to you after a while.
Or we can make a system where certain outputs from the LLM cause the execution environment fetch it data from outside sources and input it into the LLM. And then there would be model weights which make the system just sit there and do nothing, and there would be model weights which browse wikipedia all day. I think it would be apt to call this second kind a "curios" model while the pervious one is not.
I disagree on the premise that computers are able to simulate many different things; it would be just as easy to say that the universe cannot have curiosity; that the universe is just things playing out according to some preset rules.
But of course, people exist within the universe, and while our brains do function according to those rules; likely all rules that can be expressed with math and statistics; we do have curiosity. You can look at the level of abstraction of the universe, and you will not find subjective experience or choice, yet the “platform” of the universe demonstrably allows for that.
When I see arguments like yours and the parent’s, I cannot help but think the arguments would seem to apply just as well to an accurate simulation of the universe, which shows the argument must be flawed. You are a file in the universe, loaded into memory and being run in parallel with my file. If you believe physics can be expressed through math and humans have subjective experienced, then the right kind of simulation can also have these things. Of course any simulation can be represented digitally and saved to disk.
> it would be just as easy to say that the universe cannot have curiosity; that the universe is just things playing out according to some preset rules.
Behavior/dynamics depends on structure, and structure is scale dependent.
At the scale of galaxies and stars, the universe has a relatively simple structure, with relatively boring dynamics mostly governed by gravity and nuclear fusion.
When you zoom down to the level of dynamics on a planet full of life, or of a living entity such as a human, then things get a lot more interesting since at those scales there is far more structure causing a much greater variety of dynamics.
It's the structure of an LLM vs the structure of a human brain that makes the difference.
But I am saying that is the wrong comparison. The LLM doesn’t need to implement a human brain directly. It needs to implement a sophisticated enough simulation of people that the simulation itself contains ”people” who believe in themselves.
I don’t know LLMs do that, but they are excellent function approximators, and the laws of physics which allow for my brain to be conscious also can be considered some function to approximate. If the LLM can approximate that function well enough, then simulated humans would truly believe in their own existence, as I do.
And it isn’t really important whether or not consciousness is part of the simulation or not, if the end result is the simulator is capable of emulating people to a greater extent.
If you wanted to build a behavioral simulation of a human, and knew how to do it, then what advantage would there be to trying to get an LLM to emulate this simulator ?!
Better just code up your simulator to run efficiently on a supercomputer!
The LLM teaches itself the rules of the simulation by learning to predict what happens next so well.
Presumable, running a human simulation by brute forcing physics at a scale large enough to represent a human is completely infeasible, but we can maybe conceive how it would work. LLMs are an impressive engine for predicting “next” that is actually computationally feasible.
A LLM is not just a file, it also has context (the attention window in transformers).
A LLM can have curiosity in the sense that it will ask questions. It can be a useful trait as a problem often seen in current "chat"-type LLMs is that they tend to talk a bit too authoritatively about things they don't know about (aka. hallucinations). Encouraging users to give a bit more context for better quality answers can counteract this. The point could be to make chatbots not like search engines, search engines will answer garbage with garbage, a chatbot can ask for precision when it can't give a satisfactory answer.
For example:
- How much is a pint in mL?
- A US pint is 473 mL
vs
- Is it for a drink in the US? If so, a US pint in 473 mL, but in other contexts and locations, the answer can vary.
The second answer is "curious", and by requesting extra context, it tells that the question is incomplete and with that extra context, it can give a better answer. For example, a pint of beer in France is likely to be 500mL, even though it is not really a pint, it is how it is understood.
Yeah, but that's not the model being curious about things that it itself doesn't know, it'd be the model copying a human response (training sample) based on what that human didn't know.
To consider the difference, there might be some domain that the model was trained extensively on, and "knew" a lot about, but in a given context it might still act dumb and "show curiosity" just because it's mimicking a human response in that situation, not basing it's response on what it actually knows!
> Curiosity is inherently proactive. An LLM, fundamentally, is a file. It's a FILE. It just sits there until you ask for output.
And the difference between proactive and reactive here boils down entirely to an infinite loop and some kind of I/O (which could be as trivial as "web search" function call). It so happens that, in the way LLMs are deployed, you're supplying the loop through interaction. But nothing stops you from making a script and putting while(True) on top.
Put another way, if you had a brain debugger and paused execution to step it, the paused brain would also "just sit there until you ask for output". LLM interactions are like that. It doesn't make it limited in any fundamental way, much like an interactive application isn't limited in a fundamental way just because you only ever use it by putting a breakpoint in its event loop and running a cycle at a time.
Sure, but even though an LLM is only a function that "acts" (outputs a word) when called, it could (if it weren't just an LLM!) use that opportunity to act to pursue some curiosity-based line of discussion.
One limiting factor as far as curiosity goes isn't just that an LLM is a passive function, but also that it's just a statistical sequence-to-sequence machine (a transformer) - that's all that exists under the hood. It doesn't have any mechanism for innate traits to influence the generation process. All you could do would be to train that one mechanism it does have to mimic human responses judged to reflect curiosity about specific fine-tuning inputs.
I don't think that's a good analogy. We're talking about innate traits, not coarse functionality.
A plane and a bird can both fly, but a plane has no innate desire to do so, whether to take advantage of good soaring conditions, or to escape ground-based predators, etc.
An LLM and a human can both generate words, but the LLM is just trying to minimize repeating statistical errors it made when being pre-trained. The human's actions, including speech, are towards adaptive behavior to keep it alive per innate traits discovered by evolution. There's a massive difference.
An innate trait/behavior for an animal is something defined by their DNA that they will all have, as opposed to learned behavior which are individual specific.
An AI could easily be built to have innate curiosity - this just boils down to predicting something, getting feedback that the prediction is wrong, and using this prediction failure (aka surprise) as a trigger to focus on whatever is being observed/interacted with (in order to learn more about it).
> An innate trait/behavior for an animal is something defined by their DNA that they will all have, as opposed to learned behavior which are individual specific.
In that sense, most airplanes have an innate desire to stay in the air once aloft. As opposed to helicopters, which very much want to do the opposite. Contrast with modern fighters, which have an innate desire to rapidly fly themselves apart.
Then, consider the autopilot. It's defined "by their DNA" (it's right there in the plane's spec!), it's the same (more or less) among many individual airplanes of a given model family, and it's not learning anything. A hardcoded instinct to take off and land without destroying itself.
> An AI could easily be built to have innate curiosity - this just boils down to predicting something, getting feedback that the prediction is wrong, and using this prediction failure (aka surprise) as a trigger to focus on whatever is being observed/interacted with (in order to learn more about it).
It's trivial to emulate this with LLM explicitly. But it's also a clear, generic pattern, easily expressed in text, and LLMs excel at picking up such patterns during training.
> It's trivial to emulate this with LLM explicitly. But it's also a clear, generic pattern, easily expressed in text, and LLMs excel at picking up such patterns during training.
So try adding "you are a curious question asking assistant" to the beginning of your prompt, and see if it starts asking you questions before responding or when it doesn't know something ...
Tell it to stop hallucinating when it doesn't know something too, and just ask a question instead !
Honestly I don't really care what current LLMs can do, I'm more interested in fundamental limitations of AI and I think the "it's just a file" argument is nonsense and the analogy makes sense in that regard.
I think you're focusing on the wrong part of his/her "it's just a file" argument. The actual point wasn't about the file packaging/form, but about the fact that it's just passive - just a function sitting there waiting to be called, not an active agent able to act out on it's curiosity.
Curiosity is a trait of an agentic system where curiously is driving exploration leading to learning.
That depends if your goal is just to have an LLM follow a conversational style imitating curiosity (in order to appear more engaging or friendly/whatever), or whether you actually want it to BE curious about things it doesn't know, as the basis for directed learning.
You can't make a transformer-based language model "have" curiosity.
Real curiosity is an innate trait in intelligent animals that promotes learning and exploration by increasing focus and stick-to-it-ness when the situation is interesting/unexplored (i.e. not well predicted - surprising).
An LLM could be trained to fake curiosity in same way that ELIZA did ("Tell me more about your fear of prison showers"), but fundamentally it deals in words and statistics, not facts and memories, so any poorly-predicted surprise signal will be surface level such as "that was an unusual word pattern" or "you haven't mentioned that before in the input context".
Even if fake curiosity ("interest") makes the model seem more conversationally engaging, maybe helps exploration by prompting the use to delve deeper into a subject, it's not going to help learning since LLMs are pre-trained and cannot learn by experience (except ephemerally within context).