News for hackers. Stuff that matters.
Re:Generative Agents: Interactive Simulacra of Human Behavior by d--b on Tuesday April 11, 2023 @12:47AM (#35520958)
To me, having not really intelligent agents with humanlike talking abilities is the worst outcome AI could produce.

These have zero utility for humanity, cause they’re not intelligent whatsoever. Yet these systems can produce tons of garbage content for free, that is difficult to distinguish from human-created content.

At best this is used to create better NPC in video games (as the article mentions), but more generally this is going to be used to pollute social media (if not already).

Re:Generative Agents: Interactive Simulacra of Human Behavior by anileated on Tuesday April 11, 2023 @01:27AM (#35521189)
I like how this tweet puts it: https://twitter.com/FrKadel/status/1644096510357913600

The model’s designed to show “what would the answer to this sound like?”, not to provide a correct answer. Unfortunately it also 1) is profitable (look how many humans we can fire while producing kinda similar results with a tool trained on those humans’ work!), and 2) fits the age old yearning (aliens, gods) of humans for humanlike-but-nonhuman sentience, a catch-22 that’s doomed to fail.

Re:Generative Agents: Interactive Simulacra of Human Behavior by og_kalu on Tuesday April 11, 2023 @01:40AM (#35521261)
This seems hard to grasp for some reason but if the answer to "what would this sound like?" is accurate enough then it quite literally doesn't matter.

Large swaths of the brain work on prediction. You think real time reactions happen in sports? It would be impossible. You have blind spots in the eye you don't notice because the brain fills in the vision with predicted information.

If you can accurately predict what a doctor will say to arbitrary input then guess what ?, you're a fucking doctor.

Re:Generative Agents: Interactive Simulacra of Human Behavior by anileated on Tuesday April 11, 2023 @02:56AM (#35521738)
If you want to know what a plausible response to X might look like then the tool will give you that; if you are using it for getting correct information then no, it quite literally matters that the tool is not designed for that, and depending on domain and magnitude of this misapplication it could matter a whole lot.

(For example, personally, something that to me could look like a plausible answer is not exactly where my expectations are when it comes to medicine.)

Re:Generative Agents: Interactive Simulacra of Human Behavior by ogogmad on Tuesday April 11, 2023 @06:04AM (#35522962)
Your comment is verbalism; you sure you're not current-gen ChatGPT? The comment you responded to was right: If you can imitate talking like a doctor well enough, you become the same as a doctor.

Maybe we need more rigorous grounding for these discussions: We'd need to formulate a mathematical model of what these programs are trying to accomplish. Then we'd need to refer back to the exact definition of that model. Or not.

Re:Generative Agents: Interactive Simulacra of Human Behavior by TeMPOraL on Tuesday April 11, 2023 @10:46AM (#35525921)
It's literally the same thing humans do, at least to my personal experience. If you ask me a question, the first thing my mind generates is a plausibly sounding answer. That process is near-instant. The slower part is an internal evaluation - how confident I am this is the right answer? That depends on the conversation and topic in question - often enough, I can just vocalize that first thought without worry. Whether it "sounds right" is also the first step I use when processing what I hear/read others say.

If anything, GPT-3.5 and GPT-4, as well as other transformer-based models, are all starting to convince me that associative vector adjacency search in high-dimensional space is what thinking is.

Re:Generative Agents: Interactive Simulacra of Human Behavior by ChatGTP on Tuesday April 11, 2023 @08:04PM (#35533411)
It's literally the same thing humans do, at least to my personal experience. If you ask me a question, the first thing my mind generates is a plausibly sounding answer.

Ive been practicing meditation for some years now and over time I’ve realised this is not what I see happening when observing the mind. It’s one mode of operation, but it’s not the only mode. Using prediction to respond is mostly the lazy, non-interested approach, or useful if you can’t quite hear or understand someone.

What I feel lot of people have started doing is trivialising the mind. Hoping it’s all “this simple” and we’re three versions of ChatGPT away to finding God.

Maybe?

Re:Generative Agents: Interactive Simulacra of Human Behavior by TeMPOraL on Tuesday April 11, 2023 @09:39PM (#35534019)
> It’s one mode of operation, but it’s not the only mode. Using prediction to respond is mostly the lazy, non-interested approach, or useful if you can’t quite hear or understand someone.

I didn't say it's the only mode. I said it's the starter mode. At least for me, this mode is always the point at which someone's words, or my response, first enter the conscious processing level. If I'm very uninterested (whether because I don't care or because I'm good at something), the thought may sail straight to my mouth or fingers. Otherwise, it'll get processed and refined, possibly mixed with or replaced by further thoughts "flowing in" from "autocomplete".

> What I feel lot of people have started doing is trivialising the mind. Hoping it’s all “this simple” and we’re three versions of ChatGPT away to finding God.

That's one way to look at it. I prefer another - perhaps we've just stumbled on the working principle of the mind, or at least its core part. For me, the idea that concept-level thinking falls out naturally from adjacency search, when the vector space is high-dimensional enough, is nothing but mind-blowing. It's almost poetic in its elegance.

And I mean, if you believe the human mind is the product of evolution, and not a design of a higher being, then the process must have been iterative. Evolution is as dumb as it gets, so it follows that the core paradigms of our minds are continuous, not discrete, and that they must be simple and general enough to be reachable by natural selection in finite time. Of all our ideas for how minds work, transformer models are the first ones that - to me - seem like plausible candidate for the actual thing that runs in our head. They have the required features: they're structurally simple, fully general, and their performance continuously improves as you make them bigger.

Now, to be clear, my take on LLMs is that language is incidental - it so happens that text is both easiest for us to work with, and the very thing we serialize our mind states for communication. But the "magic" isn't in words, or word processing - the "magic" is the high-dimensional latent space, where concepts and abstractions are just clusters of points that are close to each other in some dimensions. I think this isn't an accident - I feel this really is what concepts are.

Re:Generative Agents: Interactive Simulacra of Human Behavior by ChatGTP on Wednesday April 12, 2023 @06:01AM (#35537284)
Evolution is as dumb as it gets

Sorry, you lost me here…we may very well find out the hard way why evolution and it’s “slower pace” is really not a bad thing. I just hope we learn the lesson in a simulator.

Re:Generative Agents: Interactive Simulacra of Human Behavior by nostromo on Tuesday April 11, 2023 @01:19AM (#35521151)
I have found Chat-GPT content to be superior to most human-created content I find in Google search results.
Re:Generative Agents: Interactive Simulacra of Human Behavior by croniev on Tuesday April 11, 2023 @01:34AM (#35521223)
It's good at presenting existing arguments in a good way. But the problem is that such models can only give back what they have seen, consolidating the status quo. There can be no reflection and no outside of the box thinking.
Re:Generative Agents: Interactive Simulacra of Human Behavior by mztwo on Tuesday April 11, 2023 @01:20AM (#35521157)
One outcome of this study was that a panel of evaluators judged the bot interactions to be more "human" than when humans impersonated these characters. So you have a point.
Re:Generative Agents: Interactive Simulacra of Human Behavior by colordrops on Tuesday April 11, 2023 @01:12AM (#35521105)
If the content is indistinguishable from human-generated, perhaps the problem is bad content, and not the source, human or not.

What do you mean by content by the way? Blog articles? Or bits and bytes in general? The right set of bits can change the world.

Re:Generative Agents: Interactive Simulacra of Human Behavior by mztwo on Tuesday April 11, 2023 @01:16AM (#35521124)
The authors of the study were clear to call out there are quite a few downsides to mass adoption of Generative Agents... the pollution and misinformation angle certainly being top of mind. I'm inclined to agree.
Re:Generative Agents: Interactive Simulacra of Human Behavior by suction on Tuesday April 11, 2023 @01:07AM (#35521086)
The key is to abandon social media and shame those who keep using it.
Re:Generative Agents: Interactive Simulacra of Human Behavior by croisillon on Tuesday April 11, 2023 @07:48AM (#35523755)
You seem to be kind of shadowban (not sure what the proper term is), you might want to write an email to the hn moderation to clear that up