AI detection is a serious problem because of scams on one side and false accusations on the other. But while building our AI detector, we realized that if people are scared or bored by your tool, they won't use it, especially when there are plenty of other options on the market. And if they don't use it, all that serious tech doesn't matter.
One of the very first decisions for our product was to make it fun and easy to use. We'd already tested different AI detectors, and all of them were very academic, clean, and serious. Not saying that's bad, but we thought, what if AI detection felt less like a security checkpoint and more like having a curious friend look over your shoulder? That's why we created this fun logo, mascot and playful interface. Later on, we though, why not use the flame from the logo and the cat from the main page as a couple of cute AI detectives? They now show up in the interface, our social media and in an animated series (I'll talk about it in the next post). They make the whole topic of AI detection feel like and adventure, not like an interrogation.
Aside from our little heroes, the interface itself also shows visual evidence of AI, inviting you to look closer, notice details, and learn something new about AI detection while checking content, instead of tons of textual explanations. We wanted it to say "hey this part of yuor content looks suspicions, see?"
All of it because we want to solve not only tech problem, but emotional as well. This way we hope people want to check if something's real or AI.
Congrats! Nice demo.
Does it work on screenshots of text, or does the text have to be copy-pasted in?
@natalie_ermishina Thank you!
Right now, text inside screenshots isn’t recognized, the text needs to be copy-pasted.
But we absolutely understand that screenshot OCR is an important feature, and it’s on our roadmap.
Scade.pro
@natalie_ermishina thank you Natalie :)
Lovon AI therapy
@olga_scry great launch, congrats! do you use it on daily basis?
@ponikarovskii thanks Anton! I do, really, especially for checking videos
I've been longing to have this tool. There's a ton of content out there that I'm not even sure of it's actually real. Congrats on the launch, @olga_scry!
Scade.pro
@olga_scry @neilverma thank you so much Neil!
App Finder
Detecting AI-generated content is clearly very important today, so I'm very happy to see a product like this.
However, in my quick tests (after upgrading to Premium), I noticed some important issues with text detection:
Test of the text detector with a text written by myself without any use of AI returned "10% of text was generated by AI" and several highlights, many red.
"x% of the text was generated by AI" suggests total certainty where there only is (not very high!) probability. The output should be something like "0% - 20% of the text was generated by AI with a confidence level of 80%". Of course, suitable test series would need to be made and published to determine an appropriate confidence interval and make the tool credible.
Some highlighted text fragments are clearly much too short to make any guess about whether they were generated by AI .
For a text generated by ChatGPT, it returned "51% of text was generated by AI". Again, several parts where highlighted for which it is clearly impossible to determine whether they were generated by AI.
The image detector worked very well in my test and returned 0% for photos I had taken, and 99-100% for realistic images generated by Midjourney (of which I couldn't have seen they were AI generated). But also here and for the other detectors,, use of confidence intervals, and publishing of tests would be very important!
Wish you success!
Scade.pro
@konrad_sx thank you so much Konrad! it's very importaint for us. i forwarded your feedback to the team and we'll get back shortly
Chatscout by Zevi
aweome product. can this be used via API?
Scade.pro
@hiya_chaplot1 thank you! we are currently working on it. Public api will be available on Q2, You can join the early access waitlist right now at isfake.ai/public-api (we don't spam in emails)
Happycapy
Scade.pro
@bartvandekooij hi Bart, yes if it's AI-generated. For example this one https://isfake.ai/explore/custom-shirt-ad-ai-model
Scade.pro
Scade.pro
@szznv thank you so much Anna, that's literally why we built isFake. and as platforms are not labelling AI content, that's a problem