Knowing how censored it gets with every upgrade and more controlled is also concerning. The freedoms are being removed and quickly.
1 Like
If you get an API key, it’s a lot less restricted. Violence, profanity, explicit language will get you an error response but it’s a lot more flexible than the free version.
If you ask the chat bot on the site to make a picture, it says it can’t. I use the text chat bot model in my Python example and it doesn’t tell you it can’t make an image, it makes an image.
software developers are going to replace everyone else with a chat bot while they talk about software developers being replaced by a chat bot … 
4 Likes
I have api keys setup with beta. Have been using Zapier and Wordpress plugins for research
1 Like
Curious moral debate could ensue…
What if you got an AI to write an article, or manipulate a chat board that could incite a riot? Convince some manipulable people to enact another Vegas attack, or shoot up a gay bar?
We start playing with probabilities, and free will becomes more and more of an illusion. It’s a classical philosophical conundrum… We can measure the “success” of an advertising campaign by its ability to coerce someone to buy a product. Couldn’t we do the same with finding the right inputs and outputs to coerce the succeptible to commit violent acts?
Do we have a responsibility to curtail potential social harm?
4 Likes
Darkness
Convince people the world is flat and they will never question anything else said again.
2 Likes
Come to think of it… If we can convince people that an invisible man in the sky cares if you masturbate, surely an AI is capable of much more…
They’d make the perfect cult leader in fact… How could you possibly argue with an entity that has scraped the entirety of human knowledge. 
6 Likes
Or a person in control of that entity.
2 Likes
Or a chatbot in control of a person who is in control of the cult. How long until AI figures out how the human brain functions and starts reprogramming us?
2 Likes
You guys should watch Westworld season 3
2 Likes
1 Like
1 billy to 10. Its gonna be all microsoft lol
Thats why musk is mad
1 Like
https://twitter.com/zswitten/status/1598380220943593472?t=3Bx7-rTSeQ9xOz4eSZLTpQ&s=19
A list of ways to bypass the filters
4 Likes
I watched Kim Possible… All the way through as a kid and one episode I’ll never forget is the robot episode when they legit start building their own hives and start manufacturing more of them that only the robots can control making them semi self-sentiment and fully reproductive via a hive.
Really fucking crazy episode and I can’t believe that it was a kids show because it was actually very well thought out and I’m guessing there was actual conspiracy theorists behind it.
2 Likes
@Future I was wrong, they hit me for $5 in damages!
I sent less than 100 requests but you’re charged by ‘token’ use and not request number. Small requests use less tokens but idk what the byte to token ratio is.
Chatgpt3 would be expensive to incorporate into a big project at this time
Abuse multiple accounts.
Have it write a script to abuse itself.
I couldn’t definitively find an answer that they’re attempting to block such attempts.
It would be a lot of work to make a new email and account then update your api key but maybe it could be done. Im more about using my powers for good these days though 
It also looks like you get $18 of tokens for free and then you pay after that, it’s not a monthly free limit like I thought. I don’t think you could rotate through say 100 API keys a month, you need a new one every time the $18 runs out
2 Likes
Mass email creation I can do easily. Mass sign up is simple keyboard and mouse click scripts.
I am speaking about a tenured ability many have. There is no verification of identity. So it is highly abuseable.
I bet people are already selling accounts.
It also doesnt ask for 2factor when you bypass by choosing to login via gmail. So gmail accounts would be ideal
For people interested in the limits of inference in models like GPT3 - and also the upcoming GPT4 -and why it is much more limited than it appears, I would suggest the book " Rebooting AI: Building Artificial Intelligence We Can Trust".
Also note you can perform simple queries that shows what GPT3 is - an algorithm to predict words that come after other words - and has no true contextual understanding of what is going on. For example:
Alex died in 1965, he was 50 at the time. Was he alive in 1940?
No, he was not alive in 1940 as he would have been only 14 years old.
8 Likes
https://m.twitch.tv/watchmeforever
Chatgpt powered endless ai Seinfeld
3 Likes