Auto-GPT: An Autonomous GPT-4 Experiment

#1

Post by keithba » Sun, Apr 02, 2023, 5:39 PM UTC

Auto-GPT: An Autonomous GPT-4 Experiment
github.com

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#2

Post by amrb » Sun, Apr 02, 2023, 6:37 PM UTC

Looks like a ReAct prompt with GPT4

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#3

Post by jadbox » Sun, Apr 02, 2023, 7:43 PM UTC

This seems... dangerously careless. What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? Sure humans also do this, but we're talking about a new level of 0day exploit carried out at scale. Sure, maybe it won't, but do you trust a hallucinating, left-brained, internet-trained intelligence to be strictly logical and mindful for all it's actions that is taking self autonomously (as this project aims to do)?

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#4

Post by roflyear » Sun, Apr 02, 2023, 7:44 PM UTC

jadbox wrote: post #3

This seems... dangerously careless. What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? Sure humans also do this, but we're talking about a new level of 0day exploit carried out at scale. Sure, maybe it won't, but do you trust a hallucinating, left-brained, internet-trained intelligence to be strictly logical and mindful for all it's actions that is taking self autonomously (…

> What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely?

How would GPT-4 make this more likely or scalable?

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#5

Post by tuxracer » Sun, Apr 02, 2023, 7:47 PM UTC

Give it access to its own controls https://git.fedi.ai/derek/talkradio-ai/issues/11

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#6

Post by sva_ » Sun, Apr 02, 2023, 7:47 PM UTC

jadbox wrote: post #3

This seems... dangerously careless. What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? Sure humans also do this, but we're talking about a new level of 0day exploit carried out at scale. Sure, maybe it won't, but do you trust a hallucinating, left-brained, internet-trained intelligence to be strictly logical and mindful for all it's actions that is taking self autonomously (…

Not sure if OpenAI has shut down any 'experiments' before, but this might be a candidate.

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#7

Post by jadbox » Sun, Apr 02, 2023, 7:48 PM UTC

roflyear wrote: post #4

jadbox wrote: post #3

This seems... dangerously careless. What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? Sure humans also do this, but we're talking about a new level of 0day exploit carried out at scale. Sure, maybe it won't, but do you trust a hallucinating, left-brained, internet-trained intelligence to be strictly logical and mindful for all it's actions that is taking self autonomously (…

> What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? How would GPT-4 make this more likely or scalable?

I'm not really picking on GPT4, but My point really applies to any LLM on autonomous Internet connected mode.

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#8

Post by rasengan » Sun, Apr 02, 2023, 7:50 PM UTC

That's pretty awesome. I wanted to try an experiment like that with ChatGPT moderating an IRC channel. I started here https://github.com/realrasengan/chatgpt-groupchat-test determining whether it was being spoken to and whether it should respond or not, and then providing that in json form so that it could be asked then to create the response in a sort of mini-chain of responses. It could also simply be asked, 'do you think this person should be kicked for their actions?' (Just for experiment and science purposes of course lol).

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#9

Post by WinstonSmith84 » Sun, Apr 02, 2023, 7:52 PM UTC

jadbox wrote: post #3

This seems... dangerously careless. What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? Sure humans also do this, but we're talking about a new level of 0day exploit carried out at scale. Sure, maybe it won't, but do you trust a hallucinating, left-brained, internet-trained intelligence to be strictly logical and mindful for all it's actions that is taking self autonomously (…

it seems to query the internet, get a response, send it to OpenAI wait ... get the response back from openAI, repeats. Any old school security scanner is 1000x more efficient, not to mention that API requests to OpenAI are quite limited.

What's eventually dangerous is that it may execute scripts on your own machine, so if it were to do some funky things, that could be rather the danger .. for yourself

Re: Auto-GPT: An Autonomous GPT-4 Experiment

#10

Post by ALittleLight » Sun, Apr 02, 2023, 7:52 PM UTC

roflyear wrote: post #4

jadbox wrote: post #3

This seems... dangerously careless. What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? Sure humans also do this, but we're talking about a new level of 0day exploit carried out at scale. Sure, maybe it won't, but do you trust a hallucinating, left-brained, internet-trained intelligence to be strictly logical and mindful for all it's actions that is taking self autonomously (…

> What if it uses the internet to seek out zero day vulnerabilities and exploit them routinely? How would GPT-4 make this more likely or scalable?

There is the possibility of API upgrades at OpenAI flipping this from "not dangerous" to "dangerous". If the AI just needs some amount of intelligence to become dangerous, then it may cross that threshold suddenly - and with OpenAI developers unaware that their next round of changes will be automatically deployed to an autonomous agent and with the auto-GPT runners unaware that new changes may substantially increase the power of the model.