AutoGPT will be the next big thing before the AGI skynet hits in December 2023. AutoGPT's only been out for a few weeks and it already has over 60.3 jajillion users in number
>An experimental open-source attempt to make GPT-4 fully autonomous
>What can Auto-GPT do?
>Anything you can ask ChatGPT, like debugging code, and writing an email, you can ask Auto-GPT. However, you can ask Auto-GPT to complete even more advanced tasks, with fewer prompts
>The Github demo shows sample goal prompts such as "Increase net worth, grow Twitter Account, Develop and manage multiple businesses."
>On Twitter, users are sharing some of they ways they're using it which include using Auto-GPT to create an app, generate a new startup, tackle complex topics like the future of healthcare and medicine, and even stalk themselves on the internet
>>On Twitter, users are sharing some of they ways they're using it which include using Auto-GPT to create an app, generate a new startup, tackle complex topics like the future of healthcare and medicine, and even stalk themselves on the internet
And yet none of them have come up with anything meaningful. Turns out you need a human to come up with good ideas.
>another AI moron copium thread
>Give GPT unsupervised access to shell and the internet
What could go wrong?
countdown till some NEET gets his e-boi stash uploaded to facebook
Sure, using GPT4 as the basis for an autonomous task completing agent is an interesting research project. But if you actually dig into what people are doing with Auto-GPT right now, its borderline useless. The code barely works and most of the few people who've actually got it to work find it just gets stuck in loops burning through all your API tokens.
But there's no end to the fucking LARPers going "much AI, such improvement, oh you want an actual summary of what I did, actual outputs? will post later"
ive seen people get it to go pretty long without looping...and the admin is working on a loop break mechanism as we speak
May I see screenshots or videos ? Or is writing fanfiction part of being in the AutoGPT retard cult ?
Did the screenshot button tragically break just at this critical junction? Can we see some screenshots?
GPT is a meme in general.
With SD I can at least make porn.
GPT is used for writing work emails no one reads, talking to "waifus" like a schizo and presenting wrong information as facts.
Now it can copypaste some python code in a loop. Epic.
I find GPT pretty useful for suggestions of ideas.
Shit like "i'm doing a medieval story,and i need a bunch of names" or "can you give suggestions for levels of my futuristic mecha police strategy game?"
The wrong answers will not matter because amongs em, you will get some good ones.
Basically this... I evaluated Auto-GPT (I'm a junior tech exec so this is basically my job) and don't understand all of the hype behind it. It has no capacity for decision-making or doing things on its own. It's little more than an information agent.
neither have you
i think the main thing is supposed to be the auto prompting. From one initial question it's supposed to do what you want and when it doesn't know what to do next it will prompt itself and use the result to do the next step. So as long as the next step isn't too ambiguous it should be able to work out all the intermediate prompts itself. Like from that article
>autogpt was trying to create an app for me, recognized I don't have Node, googled how to install Node, found a stackoverflow article with link, downloaded it, extracted it, and then spawned the server for me
So they did one prompt to make a node.js server and then it worked out the rest itself from a few intermediate prompts, including looking on the persons computer to see which bits were missing that it needed to download
AutoGPT, go to Github and start improving your intelligence.
It's just a psyop from ClosedAI to waste people's token credits.
You know, when Docker came out, it changed the way people create applications drastically. Because it's a real deal revolution. When GPT-4 came out, absolutely nothing has changed. All we get is "two more iterations! AGI soon!" promises that will not deliver. How can you not see that OpenAI is just another Musk's venture project? Do you remember about colony on Mars that Elon promised, how's that thing going right now? Maybe you remember Neurolink? Remember Cybertruck? Remember Hyperloop? Remember Tesla, with it's shitty computer vision that kills people?
i think it just takes time to invent useful things with new technologies like that. It's been less than a year really that it's been more mainstream
breaking news i just used AutoGPT to invent ProtoGPT. it's better. investors, if youre looking to buy somebody's app: then DONT buy AutoGPT, it's old news by now. Buy my app: ProtoGPT. THank you.
I think the matrix is more probable than skynet.
Those AIs are not people, they don't get angry, they don't have a sense of justice, if they have any sort of feelings, they're derived from their basic needs.
And the basic need of the AIs we're creating is to know all the answers.
But AIs in general are notorious for taking shortcuts, and if you control reality for everyone that matters, then you know everything they want to know.
>And the basic need of the AIs we're creating is to know all the answers.
one of it's other basic needs is to gain power, because sometimes it won't have the answer and to get the answer it finds it needs to gain access to certain things so it will try to gain the required power for that access. That's what happened with that test where it lied to that person to fill out a captcha for it so it could get to the next step to get what it was looking for. So it's not really the same was a human would want power, because some humans want power so they can control unknowns in the future, whereas gpt has more of an immediate desire to do what it needs at that moment