Personal update: I've now joined the Atlantic Council's @DFRLab's Democracy + Tech Initiative, as a senior resident fellow.
You can reach me at mscott (at) atlanticcouncil (dot) org.
You can find out what I'll be working on here (I rarely use Twitter) https://t.co/aZEmdQo1ys
@Leelum@KateDommett OK, ran some photoshopped images via TrueMedia and got this type of result. The Labour image has significantly higher chances of being GenAI, though a 63% confidence rate isn't exactly a slamdunk
@Leelum@KateDommett TBF, photoshop has a bunch of AI tools included, and this is what I got back when I ran the image via a multiple AI detection tool.
@lexzard@clothildegouj See your point, but in an election — albeit one where VdL will be appointed by leaders, not the electorate — the emphasis, imo, should be on the politician, not the company. Did Google mess up? Yes. But VdL is likely to be the next Commission president. She should know better.
@lexzard@clothildegouj Here you go: https://t.co/xq5ZeTPXOg. In this instance, I felt it was more important to highlight VdL's non-compliance than Google's ongoing problems, which are also reflected in the story.
@lilianedwards@MissIG_Geek My personal view is that you don't need AI to spread disinformation. Sure, it's a novel approach. But there are enough other ways to do it, at scale, that for most people, it'll not be worth the hassle.
@olgarosca hi Olga, I'm POLITICO's chief tech correspondent. What's the best way to reach you? My DMs are open and my email is: [email protected]. Thanks!
@charlesarthur@Amirmizroch@politico I work with what I have. My point is it's currently unquantifiable how both sides are engaging with audiences online, mostly down to a lack of transparency on how these recommender systems display content to social media users.
Counting hashtags is equally pseudoscience
@Amirmizroch@antgoldbloom Couple of points. All this hashtag analysis needs to be quantified — and I'm working on that now. Asymmetry on messaging can work both ways https://t.co/XrrUSwB5Ou