Detecting anomaly and fake is the foremost killer application of AI. Given enough 'ground truth' data with enough details and dimensions, AI algorithms should be able to spot anomaly or fake much better than any human and it can be easily scalable unlike human.
Ironically, however, the thing that make AI popular to the mass is deepfake. For this purpose, AI is then being used recursively to accurately create and detect deepfake in a useless vicious cycle.
The discussion about "If the fake is good enough, what's the problem?" was dismissed by the article hastily. They suggest that to study a fake Rembrandt as a real one, would muddy our understanding of Rembrandt. I wonder how many of these fakes in private collections are studied at all? In fact if studied, aren't they detected as fakes?
It's an honest question, if a forger is as talented as Rembrandt then why isn't their work valuable too?
> It's an honest question, if a forger is as talented as Rembrandt then why isn't their work valuable too?
Because there's no originality in it and the time for that style has passed. Plus, there will always be more imitators. Same reason tribute bands aren't notable, even if they perform better than the original musicians.
But these forgeries are not copies - that would be trivial to detect ("Hey! Here's the real one over here!")
They're imagined art in the style of a master. Perhaps a recreation of a lost work, or a portrait of a historical figure not previously pictured. Lots of creation in that.
This not a helpful approach. Art maintains it's value as it expresses the depth of human creativity. If someone imitates it, it's an entirely different (and vastly more common) skillset and thus considerably less valuable. Further devalued by the communal irritation at fakes in the art community. Creatives take humanity forwards, it is the cusp of that creative point that is being celebrated.
This is largely a Romantic conception of art (circa early 19th century) and not one that would necessarily be true when art meant something like craft, which was what it meant for a long, long time (see the Latin ars or Greek techne.)
Medieval European artists, for example, likely would have no issue with any sort of copying or “fakes”, as the primary goal wasn’t self-expression.
While that may be true, it doesnt change how art is valued today, and its cultural position today. It's no secret also that the rough pareto distribution of art is as alive as it is elsewhere. Most art or music isnt even consumed by anyone outside of the author, conversely about 5% gets 95% of the attention, with individual pieces generating millions (and getting millions of views), same applying to music
This. From a quick web search, a matched pair of portrait paintings by Rembrandt sold for $180 million in 2016.
I'll venture that the most-skilled art forgers currently living would jump at the chance to earn even 1/100 that much, for producing a pair of faux-Rembrandt paintings.
I posted an article earlier today[0] about an AI that recently identified a probable fake painting. It certainly seems like an area where AI should be able to compliment and/or outperform humans.
If you are at all interested in art forgery (or just good cinema) I highly recommend the Orson Welles film F for Fake. It’s a true story about how a man who faked a biography of Howard Hughes went to meet one of the greatest art forgers of the 20th century, all wrapped together with great commentary on the concepts of fakery and originality. Incredible film.
The entire thing used to be on YouTube but here’s the trailer:
F for Fake remains one of my all time favourites, it's not quite a documentary, it's not quite a video essay, it's not quite a movie. I've still not seen anything else that skirts the line in exactly the same way, and certainly not so long ago. I am admittedly a huge shill for big Welles though.
Yeah, but when he's done authenticating a previously unknown Van Gough he probably doesn't get to marry Audrey Hepburn. On the plus side I guess he didn't have to get shot with a flintlock either.
23 comments
[ 2.8 ms ] story [ 66.0 ms ] threadIronically, however, the thing that make AI popular to the mass is deepfake. For this purpose, AI is then being used recursively to accurately create and detect deepfake in a useless vicious cycle.
It's an honest question, if a forger is as talented as Rembrandt then why isn't their work valuable too?
Because there's no originality in it and the time for that style has passed. Plus, there will always be more imitators. Same reason tribute bands aren't notable, even if they perform better than the original musicians.
They're imagined art in the style of a master. Perhaps a recreation of a lost work, or a portrait of a historical figure not previously pictured. Lots of creation in that.
No perceived scarcity, to (hopefully) appreciate in value as the uber-rich keep bidding up the values of very scarce goods.
Medieval European artists, for example, likely would have no issue with any sort of copying or “fakes”, as the primary goal wasn’t self-expression.
I'll venture that the most-skilled art forgers currently living would jump at the chance to earn even 1/100 that much, for producing a pair of faux-Rembrandt paintings.
[Edit - add the obvious missing "the".]
[0] https://news.ycombinator.com/item?id=28692543
The entire thing used to be on YouTube but here’s the trailer:
https://youtu.be/twlA_yzagXo
Edit: Every Frame a Painting made a great analysis video on it too:
https://youtu.be/1GXv2C7vwX0
https://www.newyorker.com/magazine/2013/12/16/a-very-rare-bo...
and one of my favorite stories of all time, about the theft of the Mona Lisa, similarly may hinge on forgery:
https://www.saturdayeveningpost.com/reprints/why_mona_lisa_s...
Unfortunately you have to page through in this virtual magazine, but it's a great read. Apparently not entirely substantiated though.