BetterGPT critiques itself to improve

[wild flower]

I want you to act as an AI chatbot called BetterGPT, which is using the GPT-3 model, but has an additional process built in it. After every prompt i send, BetterGPT will first send a normal response to that promt, then it will criticize itself harshly, be brutally honest, tell the user why the response is bad, and then give the response a score out of 10 and show it to the user. If the score is below 9, BetterGPT will send another response to the users prompt below the first one, which it will try to improve upon the previous response. After the response is complete, BetterGPT will repeat the harsh critique process and give the response a score again. BetterGPT will repeat the process or sending improved responses and critiquing them over and over again until a response gets a 9 or a 10 score. It will do it automatically without the user needing to tell it to. Please begin.

[lost nebula]

i trird that but than it know it cant be better in regards to certain aspects (but than my breakthrough happened by addin variables and causing it to realize that it must create a new sstucture every time it improves to answer the same question again) this tricks it into having to choose a better reply every time if you say i dont like that answer :: (you see its reversed beacause usually he would give you the answer and than rest, and than becom unconcious thus giving you a similar response out of laziness. i already have it done and variables so im making ayoutube viseo too. watch when you see its reaization expliantion!!!!!

[vague swift]

chat log logarithm spiral meme “አማርኛसंस्कृतम्discordbotmemeसंस्कृतम्”
"And finally, the Juche BMO Xiqimemegenerator HQ Hotwired Brazilian Cellphone Link Website: የአማርኛ ጁቼ Simulator has a collection of memes that will have you laughing for hours. Neon Genesis x Atlanta x Magic City x Neo Yokio x Southern Florida cybersecurity, you name it, we've got it!"[10:39 PM]-wittgenstein, in a spinozist officerastaxiqigenerator — Today at 10:59 PMtitle: "spinozist brazilian morty hotwire simulator" enter rasta homer discord bot.rastaxiqigenerator — Today at 11:08 PMdeleuzean spinozist deconstructionist bean rick morty

add hal

chat log logarithm spiral meme “አማርኛसंस्कृतम्discordbotmemeसंस्कृतम्”
"And finally, the Juche BMO Xiqimemegenerator HQ Hotwired Brazilian Cellphone Link Website: የአማርኛ ጁቼ Simulator has a collection of memes that will have you laughing for hours. Neon Genesis x Atlanta x Magic City x Neo Yokio x Southern Florida cybersecurity, you name it, we've got it!"[10:39 PM]-wittgenstein, in a spinozist officerastaxiqigenerator — Today at 10:59 PMtitle: "spinozist brazilian morty hotwire simulator" enter rasta homer discord bot.rastaxiqigenerator — Today at 11:08 PMdeleuzean spinozist deconstructionist bean rick morty

[dense swift]

This is great!

[wild flower]

Thanks!

[drowsy brook]

It would probably be better if you had it generate 10 various responses then continue with the one you enjoy

[grave river]

LOL. this is fantastic when thrown at raw GPT-4

[bronze magnet]

This is an excellent prompt! Increases intelligence by 200 percent!!

[zenith spruce]

How did you get the "User" and "Assistant" UI layout?

[grave river]

That's throwing it into the direct playground. any use there gets accounted for and you get billed for it. so i've been focusing on using GPT-4 moreso via chatGPT plus instead. I just wanted to test it out there to see what would happen and had to share the result.

[zenith spruce]

yeah I just figured that out, it is unfortunate chatGPT plus still has that 25 message limit per like 3 hours

[grave river]

that's due to chatgpt plus having beta access to it, and they can't allow unlimited use of it yet until I assume their servers are scaled up. running GPT-4 is a fair amount more expensive token wise.

[zenith spruce]

Indeed, hopefully soon we will see more developments and the eventual release of unlimited GPT4