cross-posted from: https://programming.dev/post/4070845
Previous posts: https://programming.dev/post/3974121 and https://programming.dev/post/3974080
Original survey link: https://forms.gle/7Bu3Tyi5fufmY8Vc8
Thanks for all the answers, here are the results for the survey in case you were wondering how you did!
Edit: People working in CS or a related field have a 9.59 avg score while the people that aren’t have a 9.61 avg.
People that have used AI image generators before got a 9.70 avg, while people that haven’t have a 9.39 avg score.
My wife, a graphic designer, just showed me a new feature added to Photoshop that basically allows AI to take a close up portrait and turn it into a full sized photo by generating ideal content around the subject using a few simple prompts. It’s really impressive. All she has to do is touch it up to make sure it’s believable and appealing. I don’t really care, since most advertisement photography is fake anyway, but I worry how propagandists are going to exploit the hell out of AI.
what would constitute not fake advertising and photography?
you appear to have a cynical view on the matter so i assume you have a consistent and valid argument/answer.
Whoosh
I’m happy to see these results. Although I think the results for the image with the avacado is skewed as I saw that image before in an article and it may have even been the thumbnail image.
Just got a 12.
A few of them actually surprised me even after running a bunch of AI models myself
I tried to be honest about the sample images and got 50% right. I think the “drawings” are the hardest because some imperfections are intentional.
Excluding the two spoilered images, i got 13/18. A coulple I felt kinda stupid about getting them wrong after the reveal. For the two spoilered images i think i would’ve gotten the avocado one from the nonsensical highlights and shape of the speechbubble tails, but the pen drawing definitely would’ve fooled me.
I got a 9, while even knowing the two examples in this post, marking way too many as AI generated. But it got me thinking, seeing the deviantart links for the non-AI images: though maybe those are not generated by single prompts, there is a chance of heavy AI involvement in the process, one way or another. So I would actually be curious what the practical/applicable purpose of the survey was.
Maybe to add, after having read some of the linked comments, that the survey is cool and all but that’s it. It triggers people to explain their process of how they tried to spot things (as good experiment participants should), but I have only seen one comment on the purpose and methodology (as a researcher would). For whatever interesting research question, the data selection doesn’t allow for insightful answers. We can interpret these results as we want, and as we can see, people are already taking what they want from it. Sincerely, Reviewer 2
