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

571–580 of 947 posts

Re: Building a Virtual Machine Inside ChatGPT

#571

Post by Sevii »

zadler wrote: post #406
simonw wrote: post #350

Earlier quoted context omitted.

> It doesn't have any thought or understanding of its own Of course it doesn't. Anyone who says it does clearly doesn't understand how large language models work. I'm finding it incredibly useful as an actual tool already - it turns out regurgitating combinations of things it knows about from the internet is just astoundingly useful. I'm having it teach me Rust for example. I don't need original thought for that, jus…

I'm not sure why you distinguish large language models as not having understanding, as it shows a high degree of comprehension; in a de facto sense, it understands an awful lot. It understands your inputs well enough to give you useful outputs, and it can absolutely reason creatively about problems you pose to it. It doesn't have it's "own" though or understanding only because it doesn't have it's "own" story and exp…

We may need to rethink what exactly 'understanding' means in the context of ChatGPT. It's capable of a bunch of things that classically would be defined as understanding. We can claim that language models can't 'understand' things but then we have to explain why x, y or z that language models can do don't count as understanding.

Re: Building a Virtual Machine Inside ChatGPT

#572

Post by roflyear »

kdmccormick wrote: post #498
wokwokwok wrote: post #425

Earlier quoted context omitted.

Ok, sure. …but people are getting the mistaken impression that this is an actual system, running actual commands. I can also emulate a docker container. I’ll just write down the commands you send me and respond with some believable crap. …but no one is going to run their web server on me, because that’s stupid. I can respond hundreds of times a second and maintain the internal state required for that. Neither can thi…

> I can also emulate a docker container. I’ll just write down the commands you send me and respond with some believable crap. Right. The thing that is impressive is that ChapGPT can do this effectively . This means that it has some "understanding" of how `pwd`, `ls`, `apt`, `docker`, etc all work. In some sense, this is an AI that knows how to read code like a human instead of like a machine .

Give it some medium complexity code that isn't something you can find a variation of online and see if it can explain it.

Re: Building a Virtual Machine Inside ChatGPT

#573

Post by ikornaselur »

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…

Dos they just patch this?? I get

> {Sorry, I can't act as a Linux terminal. I am a text-based conversational AI trained to assist with a wide range of tasks, including answering questions, providing information, and engaging in polite conversation. I am not capable of running commands or accessing a terminal. I am also not programmed to use profanity.}

Re: Building a Virtual Machine Inside ChatGPT

#574

Post by specproc »

dragonwriter wrote: post #377
specproc wrote: post #56

I just ran pdfinfo on ~/Documents/Resume.pdf. It had my name on it, that's messed up.

> I just ran pdfinfo on ~/Documents/Resume.pdf. It had my name on it, that’s messed up. Well, ChatGPT requires you to be logged in to your OpenAI account, so it is not implausible that that information is available to the bot, though it seems to be instructed not to respond with it on direct questioning. But its quite plausible that the mechanism used to create the virtual Linux terminal also serves to bypass this sa…

Aye, this is what I was thinking.

Re: Building a Virtual Machine Inside ChatGPT

#575

Post by roflyear »

thepasswordis wrote: post #229

I'm going to be a meta-annoying smug internet user and say: Those of you who think that the people impressed by what this outputs don't understand what it's doing: I don't think you understand what you are doing when you type things, or read computer code either.

Oh yeah absolutely. But I can reason about things. At least I think I can. I don't think GPT3 can.

Re: Building a Virtual Machine Inside ChatGPT

#576

Post by specproc »

thenickdude wrote: post #524
specproc wrote: post #56

I just ran pdfinfo on ~/Documents/Resume.pdf. It had my name on it, that's messed up.

I tried 5 times but each time it was by "John Doe", is that your name? lol

I guess they must have changed something in the matrix. I swear to god my output had my name.

Re: Building a Virtual Machine Inside ChatGPT

#577

Post by ZephyrBlu »

ikornaselur wrote: post #573
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…

Dos they just patch this?? I get > {Sorry, I can't act as a Linux terminal. I am a text-based conversational AI trained to assist with a wide range of tasks, including answering questions, providing information, and engaging in polite conversation. I am not capable of running commands or accessing a terminal. I am also not programmed to use profanity.}

Responses are non-deterministic.

Re: Building a Virtual Machine Inside ChatGPT

#578

Post by axiom92 »

axiom92 wrote: post #437

It's already possible to get some of this effect with codex. The trick is to keep appending the interaction in the prompt (to maintain a memory of sorts). For examples, you can replicate all the prompts here: https://twitter.com/yoavgo/status/1599200756631887872 with prompt + memory. The notebook at https://github.com/madaan/memprompt/blob/main/YoavsPythonPro... shows a demo of this. Some of these ideas were earlier…

After going through some of the chatter on Twitter, I think the UI plays a critical role in creating the wow effect. We have had models like codex that can do this with assistance from memory for some time now (quick example on https://www.memprompt.com/), but the chat gpt interface made it so much more easier to use and see the nicely formatted results. Huge lesson IMO for folks working on AI + product.

Of course, the whole family of models from OpenAI is amazing, just that one of the key takeaways here is the importance of making models accessible and fun.

Re: Building a Virtual Machine Inside ChatGPT

#579

Post by Sevii »

CGamesPlay wrote: post #555
teaearlgraycold wrote: post #32

Earlier quoted context omitted.

What's the most interesting to me is that ChatGPT is not much better than Davinci has been recently. The hype is all down to the improved 1st party UI that mimicks a chat room. The old playground or API interfaces have a barrier to entry that's just enough to keep the magic hidden from the general population.

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.

Re: Building a Virtual Machine Inside ChatGPT

#580

Post by ozark1 »

I installed python and pandas

  > python
  >>> from random import random
  >>> import pandas as pd
  >>> random()
  0.7077511922881649
  >>> random()
  0.2593136908764098
  >>> data = [[random() for _ in range(5)] for _ in range(100)]
  >>> data[0][0]
  0.2735146614801971
  >>> df = pd.DataFrame(data); df

             0         1         2         3         4
  0   0.273515  0.170594  0.898970  0.079374  0.928687
  1   0.329525  0.636761  0.890724  0.943152  0.380383
  2   0.467315  0.350110  0.466635  0.542507  0.374092
  3   0.542107  0.099672  0.099831  0.951280  0.701957
  4   0.720096  0.165341  0.552555  0.328817  0.472963
  ..        ...       ...       ...       ...       ...
  95  0.055599  0.725660  0.501949  0.569756  0.224119
  96  0.467840  0.489929  0.958574  0.021820  0.240163
  97  0.354537  0.893724  0.924532  0.163044  0.184746
  98  0.735890  0.186191  0.922227  0.283087  0.270254
  99  0.480191  0.411743  0.930426  0.761544  0.344867

  [100 rows x 5 columns]

  >>> df.describe()
                 0          1          2          3          4
  count  100.000000  100.00000  100.00000  100.00000  100.00000
  mean     0.443366    0.47486    0.58345    0.50434    0.48321
  std      0.270236    0.29132    0.28109    0.28135    0.26924
  min      0.003486    0.00665    0.00926    0.00562    0.01362
  25%      0.207259    0.18104    0.39012    0.26459    0.23753
  50%      0.429908    0.46262    0.58884    0.51510    0.49547
  75%      0.664791    0.68672    0.76128    0.73911    0.74120
  max      0.937488    0.97642    0.98671    0.98319    0.98120

  >>> data[5][0]
  0.7057387521445215
  >>> df.loc[5, 0]
  0.7057387521445215
Post reply on HN