ChatGPT Nightmares: Scary Bears
1,762
Published 2023-11-26
All Comments (21)
-
In other words, I’m sorry Dave, I can’t do that.
-
Making an AI generorated image so scary it breaks the terms and conditions
-
Next image would have been a self portrait of ChatGPT. And that's not allowed.
-
Next image was the Care Bears, hence the copyright ©️ block
-
It's good to know that our future AI overlords will 100% know how to scare the absolute heck out of us! 😅😓😰
-
ChatGPT is way scarier than those bears!
-
AI has problems counting! Creates a picture with 4 bears instead of 3 and the next bear has 7(!) claws per foot! 😂
-
After a while I was like “Ok, that's great but it's getting a bit repetitive. Can't we just stop?”. But when ChatGPT refused to show that last imagine I got really curious of what's so scary that it's beyond the scope of the guidelines and now I really want to see it!
-
1:35 … now even it’s tongue has teeth. I like how in a previous shot in order to make it scarier, it started adding claws (I counted at least 7 on a single paw). 😅
-
I’m surprised that you got that far. Sometimes I want to tell ChatGPT that I’m a 60-year-old grown-up, for Pete’s sake. I had an argument with it about it because it didn’t like the idea of making a Motivational poster of the Shinkansen going past Mount Fuji. After a few rounds of arguing about why it violated content policy, it came out that there is a policy against using stylized fonts. It finally agreed to do the graphic part of the poster if I told it that I would put the text in in post.
-
do a pink and white unicorn next. or a purple, fluffy my pony. will ya ? will ya? will ya? pleeeeeease. or a mushroom. that would be scary: a hellscape fungi.
-
That’s an entertaining video right here! As much as I enjoyed watching this, I have to say that, what is actually scary is that chatGPT can generate images like this with no effort. I always wonder if we are not just getting way too dependent on chatGPT? Maybe not, or maybe yes… I don’t know for sure, but thanks for the video, that is a great storyline also. 👍
-
My guess is by the end, the result probably tripped some filters. Maybe it is constructing these images by using the LLM to describe the bear-a-licious details in depth to DALL-E, and then forwarding the result to you with succinct information (omitting the description it gave DALL-E) and those words either used to describe it maybe triggered the filters. Interesting either way.
-
I ran into the same problem, but with cats :p
-
If the only thing stopping it is some lines of code saying not to... what is actually possible ? Yikes.
-
Facing your fears, good therapy. 😂 this was a fun exercise.
-
I'm scared for the other more 'obvious' reason...!
-
Too spook for ChatGPT!
-
I’ve been playing around with stable diffusion to generate images some of the outputs are so realistic that it’s scary, the hallmarks I used to use to spot fakes are almost gone in the current generation of ai
-
That was un-BEAR-able. So was that joke. I suppose you could do bats or spiders or rats next. I like frogs and lizards. Should I see if you can change my mind?
