Blurring the distinction between Thought Criminals and Pre Criminals.
Here's a summary of what's been going on in the world of machine-learning, beyond what we've already covered, to kick start your week... Google ramping up AI chips: It looks like Google have hired a bunch of new chip engineers in India to crank up its efforts for building hardware for AI and mobile phone applications. The …
That's been happening since before facebook, in my experience, the sort of people who spread the real nutty stuff barely read past the first few sentences that get them enraged, before immediately reposting to spread the real truth. These posts are often accompanied by linked reputable articles, to try to give legitimacy, that mention similar words to the ranty email in the first few sentences, enough that convinces the nutty reader that the rest of the article will support outlandish viewpoint.
So to auto generate conspiracy theories, maybe all you have to do is skim scholarly articles and sites like prison planet or natural news, looking for words in the opening paragraph of the scholarly articles that overlap with the kook sites, then train the AI to use the language of the kooky site and seeding new posts using overlapped words, then publish with linked article and boom, gay frogs are coming to take over the world or something.
Koobface is an amazing vetting tool. Just trick everyone into using a website to placate their ego and boom you get to know a ton about them without making any real effort. Remember those quizzes that tell people only 9/10 get this right? I don't trust people who take those quizzes.
I've been amusing myself with https://thispersondoesnotexist.com for a few minutes now, and indeed, the mages come with occasionally rather weird artefacts. Be that as it may, by generating fictitious people you can get past the problem of having to put faces in presentations without violating anyone's privacy. Generate a few images, cut away background and presto.
That said, I second the comment about the rather small variation on glasses. And head coverings :).
Methinks that imitating AM/FM is like losing at golf; you can't just do it, you also have to not seem to be trying, which takes even more skill.
...I just realized that a poorly handmade deepfake predecessor was that film's plot device.
"Thanks, brain!" --Son Goku
"No prahblum." --Son Goku's Brain
by generating fictitious people you can get past the problem of having to put faces in presentations without violating anyone's privacy.
Anyone know what usage rights thispersondoesnotexist images are out there under? I just know I'm going to be asked if I suggest this to a few people.
I guess you have to assume that the website owns the copyright since it doesn't state otherwise. Nothing in the image properties when you download them.
I found that when you try to download the image you don't necessarily get the image you are after. when I downloaded the photo I got a different one saved than what was on screen, tried again and got a different one again. So it looks like it continues to generate them in the background even if you don't refresh the page. Even going to page info and then media in Firefox showed a different image each time.
So screen shots appear to be the only way to get a copy of the images, which suggests they don't want you re-using them.
I found that when you try to download the image you don't necessarily get the image you are after. when I downloaded the photo I got a different one saved than what was on screen, tried again and got a different one again.
I didn't have a problem saving them with Safari or Firefox on MacOS, both support "save to Downloads" via a right-click (and it saves the full size, not the scaled version it shows). Strange.
That said, I agree with another commentard that the AI should go to SpecSavers and maybe SunHut to get some more frames and sunglasses in its database :)
That's quite an interesting question. They're computer generated, so could possibly fall under artistic creations depending on the interpretation of what "art" is (and I'm going to gently tiptoe away from that debate as I'll be here all week otherwise).
That said, there is absolutely zero attempt to claim any ownership, and unless they log the creations before showing them it'll be hard to prove that an image actually originated there. Even the EXIF data in the images is about as empty as you can get, and the URL is empty too.
As a matter of fact, the website itself is simply not there - all you get is the image. There's not even a proper HTML frame around it, if you look at the page source you'll see there isn't any.
Biting the hand that feeds IT © 1998–2019