I've seen videos on walking assist devices before. It looks like they have really reduced the size quite a bit. Earlier versions really didn't look practical. I'm glad to see they are making progress.
Did I hear they were using Gwen Iffil as training model for that deep fake detection tool? Part of me is glad they’ll using more than just White Men for their training models. The other is a little 🤨 that they’re using a dead Black woman for testing their ip. I hope that they obtained the proper approval for this.
Jordan Harrod in the whitehouse! heck yeah. The DARPA deepfake detector - I sure wonder how that'll hold up to adversarial pressure. Adversarial robustness is such a tall order, there's work on it but I have to wonder whether they were able to get the theoretically-most-promising stuff to actually, like, work. and that robustness may not remain valid in the face of development of new algorithms that specifically target fooling their detector.
Thank you so much for your videos. You're my go-to when it comes to AI
I've seen videos on walking assist devices before. It looks like they have really reduced the size quite a bit. Earlier versions really didn't look practical. I'm glad to see they are making progress.
Off to watch this on Nebula! Hooray for Nebula!
this was neat thank you
i have a feeling the captions are ai generated lol
Did I hear they were using Gwen Iffil as training model for that deep fake detection tool? Part of me is glad they’ll using more than just White Men for their training models. The other is a little 🤨 that they’re using a dead Black woman for testing their ip. I hope that they obtained the proper approval for this.
Background music is way too loud compared to voices. Can’t really understand what anyone is saying
yeah like she said at the beginning. airports mannnn
edit: oh. yeah. ok. I see what you mean.
Jordan Harrod in the whitehouse! heck yeah.
The DARPA deepfake detector - I sure wonder how that'll hold up to adversarial pressure. Adversarial robustness is such a tall order, there's work on it but I have to wonder whether they were able to get the theoretically-most-promising stuff to actually, like, work. and that robustness may not remain valid in the face of development of new algorithms that specifically target fooling their detector.