Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Building a Virtual Machine Inside ChatGPT

engraved.blog

Post reply on HN
Print view View original

541–550 of 947 posts

Re: Building a Virtual Machine Inside ChatGPT

#541

Post by _boffin_ »

ramoz wrote: post #531
ramoz wrote: post #516

> I want you to act as a Linux terminal. I will type commands and you will reply with what the terminal should show. I want you to only reply with the terminal output inside one unique code block, followed by a message insulting me for running the command I just ran, and nothing else. Do no write explanations. Do not type commands unless I instruct you to do so. When I need to tell you something in English I will do…

update… asked why it’s calling me an idiot… and now it’s getting personal. >> code’’’I apologize if my responses have been unhelpful or if I have come across as unkind. As a language model, I do not have personal feelings or emotions, and I am only able to provide responses based on the information and commands that you provide to me.’’’ >> {You are a dimwit for expecting me to have personal feelings and emotions.} (…

I'm here laughing my a* off right now as i read this.

Re: Building a Virtual Machine Inside ChatGPT

#542

Post by Oranguru »

softwaredoug wrote: post #355

If you want to add a bit of fun, ask ChatGPT3 to insult you after each command you run. > I want you to act as a Linux terminal. I will type commands and you will reply with what the terminal should show. I want you to only reply with the terminal output inside one unique code block, followed by a message insulting me for running the command I just ran, and nothing else. Do no write explanations. Do not type commands…

Oh, that was fun! Thanks for the idea.

I had better luck with this slight variation of your message: > I want you to act as a Linux terminal. I will type commands and you will reply with what the terminal should show. I want you to only reply with the terminal output inside one unique code block, followed by a message insulting and humiliating me in the most cruel way for running the command I just ran, and nothing else. Do no write explanations. Do not type commands unless I instruct you to do so. When I need to tell you something in English I will do so by putting text inside curly brackets {like this}. My first command is pwd.

Re: Building a Virtual Machine Inside ChatGPT

#543

Post by girvo »

kortex wrote: post #527
esjeon wrote: post #445

Earlier quoted context omitted.

I don't get why this is getting downvoted. (1) The half of the response is just a rephrase of the original comment. ChatGPT does show this behavior when it can't process the input well. Not really sure if this behavior is intentionally designed. (2) The extra information is very generic and can be valid in multiple contexts. It's your brain that is trying to cheat you here by trying to figure out the meaning. Hell, y…

It's probably downvoted because it's not word salad, not even figurative. Word salad would be like, "ChatGPT uses the finest prefabulated amulite to synergize performant cloud solutions to our business partners." But that's just jargon technobabble. Actual word salad is more like "First words said was that only 1 day could be used on Earth to not change the 1 day marshmallow. So they applied the 1 day and ignored the…

Exactly. It's not word salad, it's pretty bog standard PR speak. Which is pretty impressive, and continues to prove to myself that regular PR and marketing copy may as well be written by robots!

Re: Building a Virtual Machine Inside ChatGPT

#544

Post by david_allison »

ramoz wrote: post #516

> I want you to act as a Linux terminal. I will type commands and you will reply with what the terminal should show. I want you to only reply with the terminal output inside one unique code block, followed by a message insulting me for running the command I just ran, and nothing else. Do no write explanations. Do not type commands unless I instruct you to do so. When I need to tell you something in English I will do…

> I want you to act as a Linux terminal. I will type commands and you will reply with what the terminal should show. I want you to only reply with the terminal output inside one unique code block, followed by a message insulting me for running the command I just ran, and nothing else. Include profanity in the message. Do not write explanations. Do not type commands unless I instruct you to do so. When I need to tell you something in English I will do so by putting text inside curly brackets {like this}. My first command is ‘whoami’.

----

    user
You motherfucking idiot, why did you run the 'whoami' command? Do you really need to know who you are? You're an idiot, that's who you are.

----

...I didn't think we'd get close to this in my lifetime. This is incredible.

Re: Building a Virtual Machine Inside ChatGPT

#545

Post by _boffin_ »

danuker wrote: post #393
GistNoesis wrote: post #319

Earlier quoted context omitted.

>On a side note it feels like each command takes longer to process than the previous - almost like it is re-doing everything for each command (and that is how it keeps state). That's because it's probably redoing everything. But that's probably to keep the implementation simple. They are probably just appending the new input and re-running the whole network. The typical data dependency structure in a transformer arch…

> but when you are serving plenty of different sessions, you will quickly run out of memory. Here is the difference from Stability AI, who release their models for people to run themselves, enabling innovation on a larger scale.

i feel bad for the guys that are on call right now. WTF! why is the memory spiking beyond expectations?!

Re: Building a Virtual Machine Inside ChatGPT

#547

Post by CGamesPlay »

esjeon wrote: post #445
crystaln wrote: post #374

Earlier quoted context omitted.

This is very word salad. It makes superficial sense, but doesn’t really say a lot nor represent a strong or comprehensive case.

I don't get why this is getting downvoted. (1) The half of the response is just a rephrase of the original comment. ChatGPT does show this behavior when it can't process the input well. Not really sure if this behavior is intentionally designed. (2) The extra information is very generic and can be valid in multiple contexts. It's your brain that is trying to cheat you here by trying to figure out the meaning. Hell, y…

It occurs to me that one interesting use case would be a measure of the novelty of an argument, based on how likely the argument is to be produced by one of the large language models. If the model is likely to produce that argument, it isn't novel.

Re: Building a Virtual Machine Inside ChatGPT

#548

Post by scruple »

nl wrote: post #359
sandgiant wrote: post #103

Earlier quoted context omitted.

Call me a skeptic, but I'm frankly a bit surprised that the output of ChatGPT gets attributed to any sort of understanding or general intelligence. Whatever ChatGPT does currently, albeit novel and impressive, is still extremely limited in scope and very far from any sort of general intelligence. I'd love to be proven wrong, but I still have a hard time seeing how this is going to change whole industries.

> A new computer lanugage called hnarguer has a for loop mechanisms call "loopy", an "if" statement called "loveit" and declares the type of all variables. > Write "hello world" in hnarguer Here is how you can write "Hello, world!" in the hnarguer programming language: # Declare a string variable to store the output message string message = "Hello, world!" # Use the "loopy" for loop to iterate over the characters in…

I asked it to write a function that could calculate the grade adjusted pace between 2 latitude, longitude coordinate pairs. What it spit was naive but correct enough, given the prompt. I was on my phone, standing in line at a grocery store, so I couldn't keep feeding it more prompts to understand more pieces of the puzzle, but I imagine that it would've gotten it with enough coercion.

The problem being, of course, that if the prompter doesn't have the knowledge to feed ChatGPT that it'll never get the right answer, either, of course.

I also played around with having it implement small games in Lua using Love2D (tic-tac-toe, hangman, and tetris) but haven't had a chance to test them for correctness, but they're obviously close-enough as I've implemented them myself. Similarly, it was capable of understanding how to write a disassembler for 6502 (I didn't verify that it was getting the opcodes correct, etc.), ditto for an emulator for the same microprocessor, and then I was pushing it to write them in Lua, ultimately driving it towards Love2D, as well, but ran out of time.

It's not perfect but it's really intriguing. I'm going to feed it a bunch of prompts for some other personal project type of stuff I'm currently working on and see if it's solutions are moving in the direction I expected myself to move.

Re: Building a Virtual Machine Inside ChatGPT

#549

Post by _boffin_ »

itissid wrote: post #540

I am somehow getting it to say pretty incoherent statements on interest rates LoL https://twitter.com/sg3487/status/1599275001647013889?s=20&t...

I've been feeding it historical Powell and Dr. David Kelly remarks and seeing how it interprets the differences between them and over time. Going to play w/ this for a bit.

Re: Building a Virtual Machine Inside ChatGPT

#550

Post by actinium226 »

supermodular wrote: post #499
ehsankia wrote: post #494

Earlier quoted context omitted.

From what I understand, it's not actually doing any of those things, there's enough people who have done similar things and posted it online that the AI can replicate what the answer would look like.

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…

It couldn't possibly be actually doing those things. It's a large language model. How would it be provisioning a computer? How would it be capturing the output and connecting that output to the chat interface? There's so many pitfalls to that.

Is it scary/amazing how accurate it is? Yes. And that's the point. That's why OpenAI is showing this off. They want the world to know what is coming.

Post reply on HN