Anyone else get the sense this is more of a Mad-Libs style fill-in as opposed to the sort of AI-generated content found in other thisXdoesnotexist sites?
As author of 'This Waifu Does Not Exist' https://www.thiswaifudoesnotexist.net/ (which does use StyleGAN+GPT-2-small, with even better anime-finetuned GPT-2-small samples coming tomorrow or so), I feel a little insulted that the 'X Does Not Exist' brand is being diluted by Mad-Libs-style efforts. The snowclone exists to show off neural net work specifically, not just any old aleatorics.
Thats understandable. Where "This Waifu Does Not Exist" serves as a demonstration of GANs, "This Startup Does Not Exist" is more of a satirical jab on how "disruptive" startup endeavors are really generic efforts that don't amount to anything
That's getting into hard to answer territory: do humans 'more or less recolor different images' when they do art, too? Many people use tracing, copy elements, imitate other artists' styles, and so on. No one learns anime art in a void. All culture is remix. And this is true of writing, as well. Consider Gene Wolfe's "The Just Man" ( https://www.gwern.net/docs/culture/1983-wolfe-thecitadelofth... ):
> From this story, though it was the shortest and the most simple too of all those I have recorded in this book, I feel that I learned several things of some importance. First of all, how much of our speech, which we think freshly minted in our own mouths, consists of set locutions. The Ascian seemed to speak only in sentences he had learned by rote, though until he used each for the first time we had never heard them. Foila seemed to speak as women commonly do, and if I had been asked whether she employed such tags, I would have said that she did not---but how often one might have predicted the ends of her sentences from their beginnings.
You could look through the various sets of samples and the more extreme psi examples and the interpolation videos and perhaps get a better idea of the extent to which the StyleGAN 'understands' faces and is being creative: https://twitter.com/gwern/status/1093701790971953152
Would it make sense to compare the output images to their nearest neighbors in the training set in order to establish some metric of "originality"? Sorry if that is a basic question, I haven't read much on image synthesis with neural nets.
Yes, and they don't usually look too much alike (see the BigGAN appendix nearest-neighbors for example of doing this) but some people argue this isn't a good enough check because the GAN could just be shuffling around patches of textures or something, which is a little goal-post-move-y for my taste.
I think "humans also do X" is fallacious. Humans also draw paintings like The Black Square by Malevich (and he wasn't the first to do it) and there's a poem consisting a single letter "i" by John Curry and another consisting of one word "lighght". Does this mean a printer capable of printing random letters is a poet and one that can produce a black page is an artist?
It's true it is hard to define what art is. But I'm pretty sure /dev/random isn't an artist, even if a human artist can produce random things.
> Does this mean a printer capable of printing random letters is a poet and one that can produce a black page is an artist?
You're asking that rhetorically, assuming people will just agree there's a difference because a human artist did it and a computer doing it isn't an artist, but you know a lot of people would be happy to take one of those forks.
Well, I guess for these people general AI already exists, or at least AI capable of producing art. Just take a random pattern, imbue it with whatever meaning you please, and treat it as artistic intelligence. With this approach, any innovation (at least as far as producing art goes) does not carry any value beyond novelty value - if /dev/random has everything one ever would need in art, then what's the difference between any new sophisticated model and just another dump of output of /dev/random?
True, but humans are generally able to start from scratch and make something. It doesn’t seem like that’s entirely what is happening here, but to be honest, it would be hard to say given the quality of the output.
I really enjoyed the videos that transition through the different options for Holo.
I've often felt that most startup landing pages are just a variation upon a theme, and this site proves it to be true, literally. They have managed to distill it down to the basic elements of unoriginality that constitute the local maxima of startup landing pages today. Personally, whenever I see one of these kinds landing pages it serves as a huge red flag.
"Many hospitals conduct nightly wealth screenings — using software that culls public data such as property records, contributions to political campaigns and other charities — to gauge which patients are most likely to be the source of large donations.
"Those who seem promising targets for fund-raising may receive a visit from a hospital executive in their rooms, as well as extra amenities like a bathrobe or a nicer waiting area for their families."
On the one hand, overpriced premium services is great way to finance core services. On the other hand, it gives influential people a distorted view of how most people experience the world. ("Doesn't even know the price of a gallon of milk.")
Also, I thought HIPAA would prevent that behavior.
Nothing in HIPAA would prevent that behavior. It does prevent hospital management from unauthorized access to your medical records for purposes unrelated to care delivery or billing. However it doesn't mention anything about access to public financial records.
Amusing, but it feels like a template where some variables are filled by random selection from an array. Closer to SimCity 2000 newspapers than to thispersondoesnotexist.com
Its address generation is a bit wonky. It doesn't respect address formats of different countries. It assumes all addresses follow this format: "<Street name> street no. <number>, <5 digit postal number>, <City>".
An actual working random address generation would be kind of cool. But this really isn't. Plus, as far as I know, postal numbers are available from OpenStreetMaps' database or Google Maps.
What if we already lived in such a world? Generated for no particular reason just for "hey look I can generate worlds randomly". I guess you should enjoy your life until the tab is being closed!
Yes, amended with the idea that with sufficiently powerful computers, simulation of worlds might be doable within things that people consider as banal as website visits.
followed soon after by a version of the Paperclip Maximizer that converts all matter into digital memory substrate to handle the constant buffer overflow of “thisthisthisthis...”
This neuron is divine. May I ask, Neuronka generates both text and icons and images or only text? And if she generates icons too, then where does she look for them, or where does she get them from, I can't believe that she draws the icons herself?
I don't get it. Is it a parody? A copy of an existing website? An attempt to see how many people will sign up without reading the URL or page? An actual startup with an "ironic" URL?
What's not to get? It's telephone calling on demand which is going to be the next youtube! I have already signed up for $89/month and sent them my resume! What a great idea. They already have three CEOs so I don't think 1 more can hurt, frankly.
The generated pages definitely are very structured, with each page belonging to a certain area/subject (ie. "Blockchain for Kids" is then presented as "Not just a Blockchain"). A lot seems to be cut-and-paste together, and not a purist implementation of GANs. With that being said, it still is very impressive that these generic page designs can feel like real world examples of what you might see in a pre-seed startup site. The satire hits right on the mark.
Business happens anywhere and everywhere. "No matter your transportation needs we can meet them with some combination of mopeds and small cars" sounds believable for a startup from any of the midsize cities in the region.
Someone made http://www.thispersondoesnotexist.com , which is an interesting site that shows how good AI is getting at generating realistic faces. It got quite some exposure.
Then parodies started appearing, like http://www.thiscatdoesnotexist.com and this one, and probably a lot of others that I didn't hear about.
