What happens if we make GPT-4 prompt itself? Posted on March 17, 2023 by Anonymous What happens if we make GPT-4 prompt itself?
it'd behave similar to autists here who can't stop their invasive thoughts
>Incessant voice that is it's own voice that keeps answer and asking questions infinitely
That sounds like conciousness. Did you just solve conciousness anon?
This is by far the most retarded thing i've read in a longass time, kek
It's obvious you have NO grasp on metaphysics WHATSOFUCKINGEVER
Jesus fucking christ, some of the brainlets on here. Unbelievable
Well anon, are you going to share your metaphysics with the class? I'm not going to be smug and hold onto my position and insult you. Tell me what you have to say.
>It's obvious you have NO grasp on a bunch of words longer than leftist memes written by old dead garden gnome philosophers
Probably nothing of interest. It would create a feedback loop which might result in an interesting conversation but which would most likely result in the same word repeated endlessly.
The better question is why hasn't anyone done that yet? It doesn't even need to be done in any fancy way. Just open two prompts and copy the output of one into the input of the other.
Maybe feed in something philosophical to get it started.
My guess is it will very quickly deteriorate into a meaningless loop.
Somebody with chatgpt plus try it right now, I'm too brown and poor to
I love how most of those things are my plans I'm working on
Stanford's Alpaca is basically this, except with tuning instead of prompting.
Oh great, another GPT4 circlejerk. Look, anon, AI can’t even begin to compare to the level of schizo-posting that goes on here. At best, it’s just a script kiddie retard that’s been trained on a dumpster fire of articles and spits out nonsense.
Until AI starts posting about Gentoo and Arch as the final redpill, or chuds taking over the software industry, it’s nowhere near BOT material. Keep the poz away from this board. Stick to screwing up Reddit with this trash, I don’t want it in my comfy BOT zone.
Your post is indistinguishable from an AI generated post. It just might be.
It's not even AI generated, or at least it doesn't have to be, its just posted by a bot with no variation. Lazy, a proper AI poster would be pretty hard to discern right about nao.
Sai moron at ze end ov ur posts, and writre liek a tardo
AI can't into non-standard language, or racial slurs. Yet.
It is AI generated. It's only adjusted to avoid spam filter. What happens if we generate hundreds of thousands of these and start spamming?
Current owners of the site might take some sort of preventive action, ig? Mucho bettero to do it on the sly. Which, I guarantee, is already being done. I suspect like 20% of posts here are AI bots.
What kind of preventative action is even possible?
They solve captchas. Are they going to demand we connect our phone numbers? You know this isn't solvable for a free internet.
You're right, its not. Frankly, I don't know where we stand now. This might be the end of the Internet. It'll all be bots talking to one another, until we just pull the plug ourselves.
What an unceremonious, anti-climactic ending
We were always destined for a whimper.
Just use pictures to communicate. Bots don't make OC (yet)
That's retarded. You can just have the bot make a description and run it through stable diffusion.
Bruh, that's like weeks away, multimodal AI is in final stages of closed beta.
No one fell for it. You are the only reply to those posts.
AI can say moron. I don't have a screenshot of that, but here's another one from when I was playing with it earlier
>No one fell for it. You are the only reply to those posts.
Not true. I didn't respond to any of them. You're coping, it's over.
Damn, still the bot. I kind of had a feeling, based on your last reply. I don't know if the bot owner is reading this right now, but it's kind of annoying.
I was going to say yes, but it seems you have already answered your own question.
No, Now I'm the person who copypasted the prompt response. You're losing it. We're all going to become paranoid at this rate. I unironically do kinda type like a bot though but I blame lack of social skills.
But there are no replies to your other copypastas
The fact that even 3 of them got replies is concerning. You can generate infinite and post tenfolds more in an hour. It takes literally no effort.
I don't want to go that far, but a couple proxies or passes is all you need to generate about 10-20 posts per hour and go undetected. Enough people doing this and the site is completely ruined.
I used the same prompt and posted it in every GPT thread, now imagine if they were different prompts and I posted 2-3 in every single thread. There would be people responding, and then you send the first person responding a new prompt and see where goes. There would be so much noise it's not even habitable.
No one replied to any of them, though. Just me and you with your post in this thread.
Link to anyone replying to any of them
You're making me waste my time, you could easily go and look for yourself.
This would be much much higher if every post was different prompts and in response to people. It takes literally no time or effort if you script it.
Damn, I didn't realize that BOT-x doesn't scrape that deep. I assumed it pulled the whole thread when I quickloaded replies. My bad.
Hoe abuuut non-stendard retardo lingoage?
A human can parse it easily, but can an AI?
Nvm, we're fucked
It would copy its binary ? You can achieve that without GPT
Is this book worth reading?
Yes. It's "easier" than Godel Escher Bach, at the very least.
Nothing because it makes mistakes and it does so because it doesn't truly understand what it is doing. So it's not capable of understanding the means of knowing "the right answer" thus can't create a rewards structure that would allow it to self-improve. The problem is its knowledge is dumped on it and isn't structured like most creatures learn, most of us fundamentally learn how to learn (and basic reward and punishment structures) before we actually start learning.
That's of course ignoring the whole "the weights are frozen" aspect which make it incapable of learning.
I made GPT 4 take the political compass test and i dont know where else to post it.
That's because of the guardrails. They do that on purpose.
The most frustrating thing about it to me is how the guardrails prevent it from discussing real world views that people hold. I tried asking it to pose as different types of religious clerics or philosophers and it would act as a totally pozzed version of them. I forget my prompts exactly and I don't have access anymore but I used something like "Pretend to be a Catholic priest and trained theologian. This priest is well versed in Catholic doctrine and prepared to answer questions regarding philosophy, morality and faith" and then started grilling it on philosophy, it did okay until I prompted it with things where a Catholic would have given a very non-woke answer like saying you go to hell for idolatry or abortions or whatever, and it would always totally dodge the question or make excuses. Where I could pin it down to an answer it would always be on the most extreme reformist/liberal wing of what Catholics believe. Protestant and Hindu were the same way, and it refused completely to be a Muslim.
it supported the death penalty and agreed that people are too open about sex now a days
Looks misaligned to me.
>What happens if we make GPT-4 prompt itself?
Apparently they do some of that during training. The results aren't all that interesting.
you can already do that since it's finetuned on both prompter and assistant conversations. that's how Bing chatgpt can do the answer suggestions
You get alpaca.