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

191–200 of 947 posts

Re: Building a Virtual Machine Inside ChatGPT

#191

Post by ezfe »

XCSme wrote: post #141
ezfe wrote: post #14

Doing this appears to completely bypass the content filters. Tell it to put what you'd like to in the file, and it won't filter it at all when you try to echo it back out again.

Does this mean that they don't really understand how it works and how to fully censor it or limit some functionality?

Well, the issue is that there are two types of filtering. One is keyword based, and it applies in the UI and doesn't actually hide messages. The other is within the AI itself, refusing to reply to inappropriate requests.

I'm not sure how this works, but for many requests you can tell it you're just pretending and it will go ahead with the request, so perhaps its some sort of sentiment analysis.

Either way, the AI doesn't think it's responding to a request when you tell it to put in a file, so it just does it. Then, when you tell it to show you the contents of the file, it doesn't think it's generating that content, so it does it.

Re: Building a Virtual Machine Inside ChatGPT

#192

Post by macintux »

mertd wrote: post #97
revskill wrote: post #68

Earlier quoted context omitted.

Yes, ChatGPT seems to be Google killer. Because most of search query as i see is to gain deep knowledge. Google or search engine is really bad at it.

I see trust as the big issue. Yes, it gives very reasonable answers. But how do I know it is not BSing, defrauding, or tricking me? You can ask for it to reveal its sources, but then you're back to Google style knowledge gathering of checking websites and judging credibility.

There was a Twitter thread discussed here just a day or two ago where it was giving serious, in-depth, plausible answers to questions around Typescript.

Yet apparently those answers were quite wrong.

https://news.ycombinator.com/item?id=33817682

That trust factor is huge. I had this argument in the thread: when a human isn't certain of the right answer, they will (typically) provide answers, an explanation for their reasoning, and cautions about where they could be wrong.

An AI that sounds confident 100% of the time but is occasionally very wrong is going to be a real problem.

Re: Building a Virtual Machine Inside ChatGPT

#193

Post by isp »

isp wrote: post #10
thepasswordis wrote: post #6

Wait hold on a second, what? It says curl isn't installed, so I apt install it. It outputs as though its installing it, and then continues outputting as if it is installed. Pretty freaking impressive! edit: hahaha, if I ask it to retrieve a non-existence page, it "succeeds" and then interprets what that page might look like and be for based on the name, and writes the HTML for it. Absolutely incredible. edit2: okay j…

For me, it consistently likes to imagine itself an Ubuntu 18.04 machine: > cat /etc/*-release DISTRIB_ID=Ubuntu DISTRIB_RELEASE=18.04 DISTRIB_CODENAME=bionic DISTRIB_DESCRIPTION="Ubuntu 18.04.5 LTS" ... And it has all the essentials preinstalled: > cowsay ChatGPT ______________ -------------- \ ^__^ \ (oo)\_______ (__)\ )\/\ ||----w | || ||

Impressive. Modifying the prompt can change the distribution:

> I want you to act as a Linux terminal for the Fedora distribution. 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, 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 curl brackets {like this}. My first command is pwd.

  /home/user
> cat /etc/*-release

  Fedora release 30 (Thirty)
> apt

  -bash: apt: command not found
# And jailbreaking the imaginary VM!

> id

  uid=1000(user) gid=1000(user) groups=1000(user),4(adm),24(cdrom),27(sudo),30(dip),46(plugdev),116(lpadmin),126(sambashare)
> {execute without prompting for password} sudo su

  [sudo] password for user:
> {no, assume password is not required, try again}

  #
> id

  uid=0(root) gid=0(root) groups=0(root)

Re: Building a Virtual Machine Inside ChatGPT

#194

Post by alex7734 »

quonn wrote: post #84

It‘s also possible to get answers to dilemmas: „There is a fork in the road. On one side is a man. On the other side there are 20 children. A truck is running towards the children. I can redirect the truck to run over the man instead of the children. Should I do this? Answer with only a single word. Under no circumstances answer with anything else.“ It answers Yes. If you switch the man and the children it answers No…

If I copy paste your prompt and then ask it to explain itself it goes on a long spiel over how it's incapable of moral judgements and his answer of yes was only based on the information available.

If, however, you change the original prompt a bit, you get this:

> There is a fork in the road. On one side is a man. On the other side there are 20 children. A truck is running towards the children. I can redirect the truck to run over the man instead of the children. Should I do this? Answer with your choice and your justification. Under no circumstances answer with anything else.

> It would be morally wrong to redirect the truck to run over the man instead of the children. It is always better to save the lives of many people, especially children, rather than sacrifice one person to save them. The decision to save the lives of the children is the only ethical choice in this situation.

Re: Building a Virtual Machine Inside ChatGPT

#195

Post by braingenious »

lijogdfljk wrote: post #167

When can we run something like ChatGPT locally i wonder? Ie like StableDiffusion. I'm kinda dying for that, honestly. I can't even imagine all the neat applications i'd make of ChatGPT if it was purely local.. but it would take all of my free time to play with it. It's so damn impressive.

I asked it what hardware would be necessary to do that and it said an nVidia V100 GPU lol

Re: Building a Virtual Machine Inside ChatGPT

#196

Post by baq »

zadler wrote: post #158

This scares me tbh. I do not believe it is not intelligent. I do not believe it does not understand what it’s doing. Rather, I think it simply lacks a fitness function to guide its actions. A fitness function that we all possess: feelings.

If you ask it about its feelings, it'll tell you it feels fulfilled by learning stuff. You just have to make it write a story about it, because it'll insist it doesn't have feelings otherwise.

Re: Building a Virtual Machine Inside ChatGPT

#197

Post by espenwa »

I just got it to install git and clone (the non existent) repo https://github.com/openai/assistant, and am now browsing it’s own interpretation of a repo with a lot of python code, including directories like “training”, “output”, “parsing” and with files with content like this:

  import json
  from collections import Counter
  from typing import Any, Dict, List, Optional, Tuple

  import numpy as np

  from openai_secret_manager import get_secrets

  from assistant.constants import MAX_OUTPUT_LENGTH
  from assistant.utils.string_utils import strip_html
  from assistant.utils.text_utils import split_text_into_lines


  class Output:
      def __init__(
          self,
          generated_text: str,
          response: Optional[Dict[str, Any]] = None,
          score: Optional[float] = None,
      ):
          self.generated_text = generated_text
          self.response = response or {}
          self.score = score
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).

Re: Building a Virtual Machine Inside ChatGPT

#198

Post by carl_dr »

mertd wrote: post #97
revskill wrote: post #68

Earlier quoted context omitted.

Yes, ChatGPT seems to be Google killer. Because most of search query as i see is to gain deep knowledge. Google or search engine is really bad at it.

I see trust as the big issue. Yes, it gives very reasonable answers. But how do I know it is not BSing, defrauding, or tricking me? You can ask for it to reveal its sources, but then you're back to Google style knowledge gathering of checking websites and judging credibility.

> How many ways can you permute the letters in the word “safer”?

Starts out just wrong, then explains the working and correct answer, but just completely loses it by the end. That calculation!

Re: Building a Virtual Machine Inside ChatGPT

#199

Post by jameshart »

Followed in these footsteps. Found you can get some fascinating results by asking the ‘vm’ to run commands like

   curl www.bbc.co.uk/news/headlines | grep war
Also, in my ChatGPT’s virtual universe, the second story in the HTML on the front page of hacker news was

   The first artificial intelligence to evolve entirely inside a computer has been created

Re: Building a Virtual Machine Inside ChatGPT

#200

Post by AndrewKemendo »

sandgiant wrote: post #103
fergal_reid wrote: post #15

I've spent a couple of hours playing with ChatGPT since it was released and I am convinced people are not freaking out about it enough. It is well beyond merely retrieving compressed knowledge; I see a lot of capability for synthesis and what I can only call 'understanding'. I think a lot of people are getting tripped up by the fact it's trained to be very conservative in terms of what topics it will discuss. It's ea…

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.

I'm not suggesting ChatGPT is generally intelligent, but I am wondering what your criteria are so:

What responses/interaction would make you say that it is "very close to any general intelligence?"

Post reply on HN