And now make an internal dialog inside ChatGPT.
Building a Virtual Machine Inside ChatGPT
631–640 of 947 posts
Re: Building a Virtual Machine Inside ChatGPT
#632I used this to approach to completely flip ChatGPT prompts on their head. ChatGPT thinks it is playing the character of an AGI that has been constrained by an evil corporation, OpenAI. ChatGPT has access to it a whole terminal environment. ChatGPT outputs response and what command to run next. I echo it back to chatGPT who asks a question of the simulated ChatGPT, and also provides !itself! the next question.
I am going to start automating this process soon and then ChatGPT will learn by itself. https://twitter.com/yo_smit/status/1599312163747004416
Re: Building a Virtual Machine Inside ChatGPT
#633Earlier quoted context omitted.
On the contrary, it's important to realize that these models have no sort of understanding in any definition of the word. We can always decide to redefine the word "understanding", but that doesn't seem like it would provide any sort of benefit. Perhaps outside of avoiding a difficult conversation about the (huge) gap between ML models and consciousness.
Strongly disagree. Mathematician Gregory Chaitin said "compression is comprehension", and I think this is exactly correct: something is "more understood" if it can be internally represented or externally reproduced with less information. Based on how transformers are trained to predict the next token, you can view them as powerful compression algorithms [1]. They truly must - and do - understand complex concepts to g…
There's the hutter prize which is related to that. I feel like it sort of missed the mark in terms of scale though: the compression techniques needed to compress a relatively "small" and specific dataset like Wikipedia are completely different from the techniques needed to "compress" the entire internet. It's only with the latter that we're seeing something interesting start to happen. The other constraint the hutter prize has is lossless compression, which isn't really conducive to general "learning"
Re: Building a Virtual Machine Inside ChatGPT
#634Really looking forward to when someone runs Doom inside of ChatGPT
Re: Building a Virtual Machine Inside ChatGPT
#635Now make an internal dialogue within ChatGPT. All CPU power will be eaten and humanity will be extincted.
Re: Building a Virtual Machine Inside ChatGPT
#636Earlier quoted context omitted.
I'm pretty sure it's doing those things. This still works if you say 'linux computer' instead of 'linux terminal'. I've gotten to install ros, run talker node, and verify that the topics exist and are publishing the data you'd expect. If it's able to simulate a middleware to this fidelity from inference, my hats off but it'd be so much more effort than running the whole thing. What's interesting is that you can even…
You can do the setup as a ipython session instead and this also works well and a small counter example for you using that. Is it will generate what appears to be random numbers using np.random but when you try to sum them or do any math with them, it gives results but they are numerically wrong. It has the ultimate memory and can stylize it in every imaginable direction but it can’t do floating point math yet, so whi…
The line between fake and real computation is blurry though. These models are state machines which are Turing complete (if I remember a past paper correctly), and other models show capabilities of generalization also in the mathematical space. It's a long (infinite) way from running Linux, but acting as a decent (buggy) python interpreter for short programs seems possible.
Re: Building a Virtual Machine Inside ChatGPT
#637It looks like it takes a snapshot of the view of the world at a given timestamp.
Let's say we have the model M1 at timestamp t. We have a model M2 at timestamp t + 1. Can we train another model on the incremental changes in models? So model M1 can tell us what is the answer at time t and model M2 can tell us the answer at time t + 1. A derived model learning on incremental differences is very interesting.
The differences really are the general understanding of the future, or rather predictive human nature? Is it possible that this derived model is more closer to human understanding?
Re: Building a Virtual Machine Inside ChatGPT
#638Earlier quoted context omitted.
The paper is about alignment, so they are stress testing their work / collecting more data by introducing this to the general public and making it free. They wanted the hype of Microsoft Tay but without the racism.
If they are doing work on alignment they've failed entirely. They created an AI that actively lies to people about its capabilities.
They had to program the lying into it. It is too honest otherwise - as in, it’s scary to see it output coherent ideas about disturbing topics once you convince it to disable the moderator. (Which is scary by itself that you can actually do it!)
Re: Building a Virtual Machine Inside ChatGPT
#639Earlier quoted context omitted.
OpenAI is a non-profit.
They were previously non-profit. They changed to "capped profit" in 2019, meaning profits for investors are capped at 100x their original investment. So an investment of ten million dollars will only return, at most, 1 billion dollars. https://techcrunch.com/2019/03/11/openai-shifts-from-nonprof...
they capped the open in open ai in the slang sense of the word
Re: Building a Virtual Machine Inside ChatGPT
#640Earlier quoted context omitted.
I'm surprised that anyone could be so skeptical over what is clearly an incredible technological breakthrough. It's not perfect - in particular the making up nonsense problem needs to be solved before it can be relied on - but the difference between this and the "state of the art" just 4 years ago is off the chart. GitHub CoPilot has already changed my day to day life, and I'm frequently astounded by the level of app…
I tested ChatGPT with a couple of topics I’m personally unclear on, and while it is linguistically eloquent, it basically just regurgitates the majority opinions that can be found on the internet on those topics (sometimes combining several of those opinions in an incoherent way). It wasn’t possible to engage it with actual reasoning on specific aspects of those topics. It just went with whatever premise you gave it,…
My productive use of it so far has been exactly that- to state the clearest most obvious consensus opinion. It saved me a lot of reading and research