One aspect of AI that gets a lot of attention is image generation. This makes sense–it’s flashy; the results are easy to see at a glance; it’s lightly horrifying because image generation is supposed to be a human thing, damn it! Here’s an example from a popular project. This person does not exist:
The latest craze in this area is text-to-image. The neural networks for this are evolving rapidly, and lately have been producing some pretty impressive results.
VQGAN + CLIP “painting of small cabin in the middle of snowy mountains in the winter at night in the style of disney trending on artstation | unreal engine” pic.twitter.com/NORfb5ZEIg
— AK (@ak92501) June 19, 2021
How does this software work? What can its eccentricities tell us about AI? About the future? Join me below the fold for a discussion of Generative Adversarial Networks, or GANs…
The simplest GAN is actually two neural networks pitted against each other. One, the generator, creates the images; the other, the discriminator, says when the images are ready. Seems simple enough. But how is this network trained? Let’s use the This Person Does Not Exist GAN as an example.
We begin with a dataset of head shots and two blank-slate networks. Then… it’s a bit involved. First we train the discriminator to identify pictures of head shots, just like you’d train any other such network. Next, we have the generator make some images, and train the discriminator further by telling it that these are fake. An earlier post gets into what this process looks like. It is as straightforward and as well-understood as these things come. And then we flip the script, having the generator create images until it can fool the discriminator into thinking they’re head shots.
We repeat this whole process again and again and again. At the end, we have an AI that can generate head shots.
As is common in this field, narrow-purpose networks tend to perform better. You can compare the two outputs above: the dedicated head shot generator creates images that are almost indistinguishable from the real thing, at least without close inspection. Conversely, the more general text-to-image network, called CLIP + VQGAN, is impressive for what it is, but obviously has a long way to go.
One interesting thing about CLIP + VQGAN is how much better it works when you find the right magical incantations. In the tweet above, you can see that the prompt is “painting of small cabin in the middle of snowy mountains in the winter at night in the style of disney trending on artstation | unreal engine”–this is because adding “trending on artstation” and “unreal engine” steer the network in a more painterly and recognizable direction. Working with it is like trying to extract the right wish from a deaf genie, to paraphrase some tweet I saw. This is because the guts of a neural network are ineffable; it attaches meaning to things in a way that we do not, or at least, in a way that it cannot tell us about. (I’ve covered this in some other posts on this topic, but I love this as an illustration.)
So, where do we go from here? Obviously CLIP + VQGAN won’t be taking over the art world any time soon, but you can see how it makes a great resource for visual artists, just as text generation can be a great resource for writers.
Miscellany
Here’s another good one:
This brought some flashbacks … “child drawing of nuclear war” #MachineLearning #aiart #VQGAN pic.twitter.com/BFSRsf4JYM
— david gallay (@svengali) July 16, 2021
#LatentVisions is another network that’s doing pretty impressive stuff.
Just for fun, I also asked the text generation AI GPT-3 about GANs. Seed text in bold:
What is a GAN? A GAN, or Generative Adversarial Network, is a machine learning model that is used for generating realistic looking images, audio, and text. GANs can be used to create more convincing fake data for training a machine learning model. GANs consist of two neural networks, a generator and a discriminator. The generator attempts to learn how to output realistic looking images. The discriminator attempts to learn how to distinguish between real images and fake ones.
What are the benefits of GANs? GANs have several benefits that make them attractive to researchers and companies. They can be trained with unsupervised learning, and they can generate new images based on existing ones. They can be trained with unsupervised learning, and they can generate new images based on existing ones. They can be trained with a smaller dataset than required for generative models.
What are the drawbacks of GANs? GANs are computationally expensive, requiring two neural networks (one generator and one discriminator) to be trained. They also suffer from unstable gradients and mode collapse. The recent work in this area is interesting and worth looking into.
Not bad!
West of the Cascades
Once Bitcoin is banned, all those mining computers can be repurposed! Pretty interesting and slightly unsettling stuff.
VeniceRiley
Great. Make me an audio conversation of Barack Obama asking Hillary Clinton to send more baby blood in the style of cell phone audio.
Q conspiracies will get amazing
Van Buren
I keep reading that as VOGON and marvelling that their art is so much better than their poetry.
Old School
@VeniceRiley:
If you want Q conspiracies to get amazing, have Obama claim that Mike Pence finished the last of the previous batch.
citizen dave
Interesting!
Will we eventually get to the place where each (first we world) will be able to imagine any type of art, movie, song, etc. A bit off topic but as a music fan I’ve wondered if there will come a time when you can have an artist play any song in any style etc. And any combinations of bands.
Or will Idiocracy prevail?
theflex
Great, but why? No way that the bad uses of these technologies won’t outweigh the good.
MattF
This is a science thread… so I can link to this astonishing post from Derek Lowe describing how plants were ‘persuaded’ to produce a human enzyme that greatly increased the plants’ productivity. I really got that ‘IT’S ALIVE’ vibe from this one.
NotMax
The perfect stocking stuffer! Hours of fun for the whole family! It’s “Infinite Number of Monkeys in a Box.”
By Marx*.
*that last gratuitously thrown in to tickle the olds’ nostalgia bone
dm
The latest XKCD: https://xkcd.com/2494/ is targeted at GANs.
Elizabelle
Worried. Bob Odenkirk was hospitalized last night in ABQ after collapsing on the set of Better Call Saul’s last season. And no news since.
And just checked TMZ for any update, and Dusty Hill, ZZTopp’s bassist, has just died. Damn.
Sorry to be the bearer of bad news here.
The Moar You Know
The fake young lady above, well, that level of fakery has been around for a while and one area in which it’s being used in is, of course, ‘simulated’ child porn. Which is not illegal in the US.
So far, detailed digital forensic analysis can show the difference. The human eye cannot.
Courts are going to get a workout with that shit.
WhatsMyNym
At first glance the photo of the girl looks realistic. At second, it just looks creepy.
Oklahomo
https://thiscatdoesnotexist.com/
Martin
Things like photo generation are interesting, but not terribly helpful. It does more harm to destroying people’s trust in photographic evidence.
I think generative design – using ML in CAD + simulation software like Ansys + additive manufacturing is going to be revolutionary. We train engineers to design around conventional tooling, which additive manufacturing frees us from, but the new freedom becomes too complex for traditional determinative design to optimize to, and the traditional design, build, test cycle is too slow to iterate at a large enough scale to get there. But you can do that very easily with generative design – modify a design, simulate it, pull the parameters back into the ML model, repeat. Then, produce a physical version it it, test it, and if you need to refine your parameters, odds are the model already stumbled on something close to it, and can start much closer to the target.
This is also being done in materials science in a few ways. One approach is mining through published papers looking for materials and observations that were overlooked, then feeding that into the next approach which models the process of creating new materials and predicting their properties.
The big one, though, will be for code generation. And I think for code testing. I’d love a service I could point at my code and ask it to break it. Basically, find gaps in my unit tests and input validation, integration tests, etc.
Spanky
@NotMax: Hasbro or gtfo.
mrmoshpotato
@WhatsMyNym: Something about the smile just seems – off.
mrmoshpotato
@NotMax: Sir! I wish to invest in your business!
mrmoshpotato
Siri, draw the gotting of a basketball jones, ooo baby, ooo ooo ooo.
Oklahomo
Also, this one is easy to play with if you want to create cheesy art: https://deepdreamgenerator.com
Baud
A Baud! 20XX! voter!
Major Major Major Major
@The Moar You Know:
Not sure why it should be.
Baud
@The Moar You Know:
It’s illegal if it meats the legal standard for obscenity, in places where obscenity is criminal. It’s not illegal as child porn.
Enhanced Voting Techniques
VQGAN or Vogan?
So we now have a Thomas Kinkade simulator. How long before Science gives us the first NYT editorial simulator?
Steeplejack (phone)
@Martin:
What is ML?
Major Major Major Major
@WhatsMyNym: this particular GAN struggles with teeth and ears
?BillinGlendaleCA
@Enhanced Voting Techniques:
How would we know the difference?
Martin
@Steeplejack (phone): Machine learning. Umbrella term which GAN is a specific implementation of.
NotMax
Recently declassified: primitive example of AI-designed meatloaf.
“Boss! We’ve got the result! Whaddaya think?”
“Well, that’s a solid 16 weeks of number crunching shot to hell.”
:)
Omnes Omnibus
@Major Major Major Major: It has been litigated. Too much like criminalizing thoughts. Actual physical child porn requires the abuse of a child whereas a fictional representation does not.
TriassicSands
Damn, and I got my grandson a date with her on “DateImaginaryPeople.com”. Will he ever be disappointed!
Enhanced Voting Techniques
The answer is complex.
Martin
@Major Major Major Major: It raises a really interesting question. Does the law exist to protect victims, which there isn’t one of in this case, or does the law exist to enforce norms such that it doesn’t matter if there is a victim or not.
Similar problem we have with vaccine disinfo. The ‘victim’ is society, collectively, but nobody individually. But we overindex on individuals rather than society. So society has no rights, but the dipshit who thinks the vaccine makes you magnetic does.
Major Major Major Major
@Martin:
There’s also a harm reduction angle: does this content act as a ‘release valve’ for the behavior we’re actually trying to prevent?
Enhanced Voting Techniques
We’ve been there since the 90s easily. It’s not really a stretch to create a decent fake person in photoshop and heck, consumer level 3d like DAZ can do stuff that looks like the real thing now with some touch up from a talented artist. The real test would be if the AI can create the same person in a series of images doing some activity. Just for example Hillary Clinton selling babies to Space Lizards, since that’s the kind of stuff it is likely to be abused with.
One thing that will give the fakes away is programers being a mostly dude bro profession; inevitably and without exception all the women in these images created by these AIs will be unnaturally busty as shown by Deviant Art.
piratedan
@Baud: just another instance where the Dem’s are behind the GOP, because Trump has been using these “people” anecdotal examples for at least 5 years now…. Granted they always seem to be crying but perhaps that what happens with beta rollouts…
Major Major Major Major
@Enhanced Voting Techniques:
isn’t this basically a deepfake?
Hilbertsubspace
I’ll be more impressed when they design an AI that can run a Dungeons & Dragons campaign just from audio inputs. Parsing natural language is “Hard”, but dealing with gamers is even harder.
I also thought of Vogon poetry when I saw VQGAN.
Ken
Pipes the request “painting of small cabin in the middle of snowy mountains in the winter at night in the style of disney trending on artstation” into Google Image Search and returns the top hit?
NotMax
Sixty years ago in computerland.
It was released as a record on a mini-sized platter. Whether it was a promotional item or whether one had to visit a record store to purchase it, don’t rightly know as I received it at the time as a gift (I still have it.)
Ken
@MattF: Building a better triffid with Science!
Steeplejack (phone)
@Enhanced Voting Techniques:
This problem has been solved.
Roger Moore
@VeniceRiley:
You aren’t thinking big enough. It should be, “Make me a crazy conspiracy theory targeting Hillary Clinton and Barack Obama in the style of QAnon”.
Enhanced Voting Techniques
Yes, and I thought you were warning us to be ready for it. I can see tech being used animation, but it’s almost certain to be abused.
Major Major Major Major
Somebody should set up a bot that generates VQGAN paintings from GPT-3 prompts and sells them as NFTs
zhena gogolia
Too bad this political blog has no political posts except in the middle of the night.
NotMax
@Enhanced Voting Techniques
Silly me, I thought the idea was to create something wholly original, not just mimic reality.
“Is there a discount for paying in quatloos?”
:)
Roger Moore
@The Moar You Know:
A new use for AI generated photos is as the avatars for social media bot accounts. The bots have been caught too many times using reverse image search, so now a lot of them use completely made up images.
Enhanced Voting Techniques
What, like Trump showing affection for Don Jr? I would think that would be asking to much for even paired super computers.
Roger Moore
@Enhanced Voting Techniques:
It’s already posting on Twitter as @DougJBalloon.
Major Major Major Major
@zhena gogolia: don’t look at me, I’m on the tech and lifestyle beat
Elizabelle
@zhena gogolia: I know. “And the portions are too small” too. //
Maybe the trick is to go back to the morning thread.
Or do something else.
Major Major Major Major
@Enhanced Voting Techniques: yeah we’ll probably be drowning in this stuff one of these days.
Bill Arnold
@WhatsMyNym:
Pretty much. With this generation of facial image generator at least, you can (often) tell with those parts of your brain/mind that detects creepiness/incongruity (whatever they are, however they work). (There are a few sites where one can test oneself at this sort of real/unreal discrimination. )
It takes a little practice and some mental effort, though.
And I’m certain that generators that get around such human discrimination (e.g. by using humans/mechanical Turk as the adversary) could be built. (More interestingly, an analysis might tell us more about how the uncanny valley works in humans.)
NotMax
@zhena gogolia
Full service blog. Can do the Doom Boogie only so many times on the dance floor before needing to sit down and catch one’s breath.
;)
Omnes Omnibus
@zhena gogolia: You can always ask for a refund.
Old School
@zhena gogolia:
Major Major Major Major
John Revolta
Handwriting’s on the wall for the modeling industry……………………
Elizabelle
@Old School: I am with zhena on this one.
If we don’t have anything to say about Simone Biles or AI, where do we go? I am not a fan of two “specific” threads, one after the other.
NotMax
@NotMax
Amend that (no edit function).
Full service blog. Can do the Doom Boogie only so many times on the dance floor before needing to sit down and catch its breath.
;)
WhatsMyNym
@Major Major Major Major: It just looks like a face put together from different people of different ages. And then given a soft focus and strange backlighting. The hair is all wrong.
ETA: the ear stud just looks like it was added by a first time user of photoshop.
Old School
@Elizabelle: Talk about politics in the comments anyway?
Enhanced Voting Techniques
@Major Major Major Major: snark assiside- got any stuff the AI doing old, out of shape or otherwise not beautiful people? That’s usually when this stuff falls apart because of the obsessions of the people creating the software.
guachi
Fake images like the one posted have a weird softness to them. The image should be sharper in some way but it’s not. That’s the only way I can tell the difference.
mrmoshpotato
@guachi: Does the ear seem too far back on the head to you? Or is it just me?
NotMax
@Enhanced Voting Techniques
Not widely known but Jabba the Hutt was voted Mr. Beefcake three years running on his homeworld.
zhena gogolia
@Major Major Major Major:
I don’t blame you!
Major Major Major Major
@Enhanced Voting Techniques: You can keep refreshing https://thispersondoesnotexist.com/ to get an idea of the image space. I assume it’s trained from published head shots, so you’ll have a natural bias towards the sort of head shots people publish.
zhena gogolia
@Elizabelle:
It’s funny, because I don’t go to any other blogs, and I guess blogs are a dying breed anyway. So I’m reduced to checking various twitter feeds every few minutes. (I just happen to be taking a day off work today, so I was looking forward to luxuriating in BJ.)
I’m watching the 1/6 testimony again, because I only got it in fragments yesterday. Hodges’s testimony is shattering.
zhena gogolia
@guachi: She has a weird, “Chucky doll” look in her eye.
Major Major Major Major
@zhena gogolia: Well that makes one of us!
Baud
@Ken:
Magnets.
Elizabelle
@zhena gogolia: Kay is in the “kids on vents” thread. Yea! She’s talking about sending kids back to school, and how important it is.
Head there, and we can figure out some other topics to bring up. (Other than a quick scan of the news, when I found out about Bob Odenkirk and the late Dusty Hill of ZZ Topp, I’ve not been reading any news today.)
Bill Arnold
@Major Major Major Major:
@ctrlcreep (twitter) had a couple of their (often intense) magical realism tweets given the treatment (these link the original tweet)
Old School
@Major Major Major Major:
There’s a 25 minute video about the process here for anyone who would like to go more in depth. (Disclaimer: I didn’t watch it.
Edit: Shorter video here.
Baud
@zhena gogolia:
We need an AI to generate more timely political posts.
Jim, Foolish Literalist
I’m old school. I still blame Obama
Elizabelle
@zhena gogolia: Yeah, I need to watch the whole thing from the beginning, too. What I saw was, as you say, stunning.
I have a nephew in law enforcement in the DC area. He could have been any of those testifying. Or the late Officer Brian Sicknick.
I want a deep dive into what brought us January 6th, including the media-sphere. It’s not just score-settling. We need to know.
trollhattan
Programming note: NBCSP currently showing Olympic dancing horses. I repeat, dancing horses.
Or, watch on regular NBC, USMNT dismantling Iran in hoops, like Allah and Dick Cheney intended.
zhena gogolia
@Elizabelle:
Yeah, but it looks as if it’s going to be a Sinema thread now. No energy for that either!
I guess I’ll go back to Richard Armitage videos.
Bill Arnold
@WhatsMyNym:
And gender presentations. Categories are being mashed up that should not be mashed up.
NotMax
Guess it’s worth tossing in again. You wanna talk spooky computer imagery? The many faces of Abe.
Also too, this compilation.
Elizabelle
@zhena gogolia: Find another topic and put it in the “vents” thread. Something will take. I don’t know much about Sinema either, and not in for gloom and doom either.
germy
Paul McCartney’s latest video features a deep fake version of his younger self, portrayed by Beck.
https://www.youtube.com/watch?v=t2Q4hAQ4fEw
germy
@Elizabelle:
I don’t know much about Sinema, but I know what I don’t like.
Jim, Foolish Literalist
@germy: Whoa
Am I the only one who would’ve thought of Beck as two-thirds John, one-third George and a splash of Ringo?
NotMax
@Germy
Looks more like a young Zachary Quinto than McCartney.
Delk
RIP Dusty Hill.
germy
@Jim, Foolish Literalist:
When I first saw the video, I thought it was a de-aged Paul, but the dance moves reminded me too much of Beck.
This technology will create a lot of mischief in the wrong hands. Every republican campaign ad will feature democrats doing and saying bad things.
germy
@NotMax:
Spock or Greenwald?
NotMax
@germy
Greenwald goes with the meat, Spock with the fish.
;)
Martin
@Major Major Major Major: Right. And we can’t forget all of the ‘norms enforcement’ we did against people of color, the LGBTQ community (which we are still insisting on doing) etc.
That’s why we need to ask the question. Often. And be honest about what job we are actually trying to achieve.
Another Scott
Interesting stuff. Thanks.
I’m reminded of those electromagnetic spectrum graphs from DC to gamma grays and oh what we can detect with our senses is this teeny tiny region right here. Presumably AIs that aren’t trained with all those similar subtle details of what they’re trying to mimic will still be able to be falsified without too much trouble.
[insert philosophical discussion about whether an AI that can satisfy arbitrary consciousness test is really conscious]
Cheers,
Scott.
Elizabelle
@Delk: Yeah. 72 is young these days. In his sleep, allegedly.
Ken
Well, if we’re getting philosophical, does that mean the AI would have all the accidents of consciousness, but lack the substance of it?
Elizabelle
@germy: I don’t know much about AI, but WRT image manipulation, it has the potential to be even more destructive than social media. You won’t be able to believe your eyes or ears, unless you see something in real time or real life.
It all becomes an assault on the truth, and we do not live in a virtual world. We live in a real one. That is warming. Rapidly.
I am fucking sick of the tech bros who cannot think beyond their next toy or round of capital.
The question is not always whether you CAN do something, but whether you SHOULD.
Of course, the designers do not have the big picture, and they want to beat their competitors.
People already have really fragmented attention spans. More circuses, please!
germy
Agree.
Elizabelle
We already have an enormous problem with people in this country who know/believe things that are not true. I speak of those who get their worldview from Fox News and rightwing media.
And now we need to add altered images all over the place? Plus deepfake voices?
How does any of that serve us?
I think this has the potential to be really monstrous.
moops
NY Times pitchbot will become sentient and still hate us all.
Jim, Foolish Literalist
Since this is a technology thread: Is this true?
The microchip shortage is tied to GCC? I thought it was part of a broader Covid-related supply chain fuck-up?
Geminid
@zhena gogolia: I saw that Chuck Schumer says he will bring the bipartisan infrastructure package to the Senate floor this evening. So there should be good politics post about that, maybe the middle of tonight.
Another Scott
@NotMax: Neat stuff. Lincoln’s eyes look weird in most of the first link, but maybe his eyes were weird.
I’d forgotten that he was only 56.
:-(
Cheers,
Scott.
moops
@Ken: Having all the accidents of consciousness is likely the same thing as consciousness. Evolution whipped up consciousness from pond scum+time. It is not really a designed thing.
Jim, Foolish Literalist
@Geminid: five minutes after a thread about the Olympics and five minutes before a “respite” thread about, oh, let’s say blue-footed boobies
Another Scott
@Ken: Ouch. That makes my head hurt.
:-(
Thanks. ;-)
Cheers,
Scott.
moops
@Jim, Foolish Literalist: the chip shortage is unrelated to COVID. It is also only lightly related to climate change.
the chip demand curve continued to be exponential, and it turned out there are raw materials and manufacturing capabilities that only grow linearly, at best. The curves finally crossed sometime last year.
The chip supply problem is sort of the canary in the coal mine for what capitalism looks like when it hits hard finite supply and some of your raw stock is not fungible.
trollhattan
@Jim, Foolish Literalist:
Weird things happen sometimes. I can recall a time [puts feet on coffee table] after some Japanese quake or other a factory that produces digicam sensors was knocked out for months, affecting new camera production. Same quake, different factory produced display panels used by BMW and some models (notably Minis) were unavailable with nav systems.
Ya just never know.
Old School
@Jim, Foolish Literalist: Bigfooted by John Cole’s “Here’s what I canned today” post.
Another Scott
@Jim, Foolish Literalist: There’s apparently some truth to it.
DW story (from May 19).
Lots of water is recycled/filtered/treated in modern plants, but lots is required just the same. But other stuff is going on.
(I’m still surprised how little the Texas winter power fiasco is talked about given all the semiconductor plants in Austin and other places in the state. NXP makes lots of semiconductors for cars.)
Cheers,
Scott.
Gin & Tonic
@Elizabelle: I believe I’ve referred before to a book on my shelf entitled The Reconfigured Eye. The subtitle is “Visual Truth in the Post-Photographic Era.”
It was published 30 years ago.
Cmorenc
@WhatsMyNym: i will bet you saw nothing creepy whatsoever about the image of the girl until you were informed the image is a completely fictional girl generated by AI. It is precisely that this fake is so convincingly free of any clues of fakery (except perhaps to true experts) that it is so creepy – it undermines your faith in your ability to sense reality from fiction plain before your eyes.
Elizabelle
@Gin & Tonic: Around the time of “Amusing Ourselves to Death,” perhaps.
VeniceRiley
The harmful use of this will not be a big complicated production. It will be fakes with a short turnaround time that can be deployed , gone viral, and mass accepted before the truth can get its socks on. That’s why I have the audio only example.
But that photo is amazing because it not only looks real, it has me thinking of who it resembles because the features feel so familiar.
OFFTOPIC My fiancé just got a brilliant Nuka Girl tattoo….
Elizabelle
Cruising TMZ for any updates on Bob Odenkirk … and now Ron Popeil has died. Aged 86.
Is there a word for not mourning someone’s loss because you thought they’d already been dead for years?
Anyway, RIP with your ginsu knives or whatever it was you used to sell …
laura
@Elizabelle: You said it better than I could. It’s deeply unsettling. And it raises the issue of whether Code is Law. And there’s always this:
https://youtu.be/yWPyRSURYFQ
dm
…Plus real horrors committed by TFG or whoever the current cult leader is will be dismissed as “deep (state) fakes”.
Ken
“I had a brief feeling of vigoda”?
Elizabelle
@Ken: Vigoda. LOL.
laura
@Ken: “I had a brief feeling of vigoda”?
Nominated for rotating tag!
Bill Arnold
@Cmorenc:
You can choose not to believe me, but I get about 90 percent correct on https://www.whichfaceisreal.com/index.php (and get annoyed by my mistakes).
Subjectively (to me), it seems like there are hidden categories being devised by the network then combined even though the combinations don’t make sense.
Bill Arnold
@dm:
That’s one of my worries too; that it provides deniability cover for malignant activities.
There needs to be a market for certified video/stills; a tamper-resistant camera, maybe added to a blockchain within seconds to provide an unforgeable time stamp for the original evidence.
(You might occasionally see SHA256 strings in tweets on twitter; this is a low-rent way of establishing possession of a block of data at some time point.)
Perhaps also media outlets (“trusted”) that make a point of severely punishing forgeries and false claims of forgery.
J R in WV
@Elizabelle:
Billy Gibbons and Frank were touring with their former guitar technician on bass recently. I assume that was OK with Dusty, they seemed to be good friends for 50+ years now…
ZZ Top is scheduled to appear here in the Civic Center arena, where I saw them some 6 or 8 years ago. IT was a great show…
RIP, Dusty, what a great bass player he was.
Roger Moore
@Major Major Major Major:
It’s interesting. I noticed a few things they seem to get consistently wrong:
WhatsMyNym
@Cmorenc: I’m looking at this on my phone. What stood out was the eyes and shadows under them. Then the weird ear stud. Looking at it more closely I saw the other problems. It’s a fake and or a real bad photo.
Yes, I’ve done some photography and worked for a professional. Back in my day, we could do a lot with makeup, lighting, and then finish it up with tricks in the lab.
guachi
@moops: The chip shortage is partly tied to COVID because demand for things that used chips increased because people were stuck at home.
The shortage is also being driven because Intel’s fabs have low yields. The best fab in the world, TSMC, has multiple customers (Apple and AMD are TSMC’s two largest customers with Apple 25% all by itself) producing quality products that are in high demand.
Apple’s iphones? TSMC.
The new Sony and Microsoft consoles? TSMC (made with AMD parts)
Roger Moore
@Bill Arnold:
I also do well on the “which one is real” test, but it’s much easier when you know in advance one of the faces is real and the other is fake. I’m pretty sure I would do substantially worse if I were presented with single pictures and asked if they were real or fake.
Jim, Foolish Literalist
Another Scott
Relatedly, …
(via IamHappyToast)
Cheers,
Scott.
Major Major Major Major
@Bill Arnold: “added to a blockchain within seconds” is a bit of a contradiction
Mary G
Elizabelle
@Jim, Foolish Literalist: Oh, thank dog!
The no news for so many hours was worrying me. Wishing Saul/Jimmy a full recovery.
Tweet from Dave Itzkoff:
Elizabelle
@J R in WV: ZZ Topp scheduled for Richmond in September. Had not realized it was just 2 of the original big 3.
I got to see them at the Hollywood Bowl, probably 10 years ago. They were opening for Tom Petty.
Excellent all around, both groups. RIP, Tom and Dusty. Too soon gone, and both out at the top of their game, for age.
Roger Moore
@Major Major Major Major:
It’s only Bitcon [sic] that insists on a long time between blocks. Some of the other cryptocurrencies have a new block every few seconds, e.g. Ethereum adds a new block about 5 times per minute.
Bill Arnold
@Major Major Major Major:
There are permissioned blockchains that are faster, and are largely a consensus mechanism supporting immutability. (At least immutability unless there is substantial collusion between member parties.)
e.g. https://hyperledger-fabric.readthedocs.io/en/release-2.2/whatis.html
(Not claiming expertise, though.)
Bill Arnold
@Roger Moore:
And it’s annoying to even have to worry about it. It requires extra mental energy to evaluate images (or soon, video) for fakeness, akin to the mental energy required to appreciate visual artwork by talented artists, but is only needed because a-holes are polluting our information spaces with unlabeled fakes.
Just Chuck
The girl pictured has an “off” look as do many of the other images generated by that site, but going by a dozen or so reloads, the majority of them looked absolutely real.
One thing I’ve noticed from animation like the ultra-realistic episodes in Love, Death, and Robots is that even with perfectly realistic models, motion always looks off by just a wee bit. But of course kinematic models will continue to get better too.
There are those who call me...tim... (Still posh)
Thanks. I wasn’t terrified enough already.
John Revolta
@Ken: Hilarious.
I saw Abe Vigoda on the subway in 1985. We gave each other the nod…………..when I got to where I was going nobody believed I had seen him because people already thought he was dead THEN!
different-church-lady
No, it’s not lightly horrifying. That image is ENTIRELY horrifying.
different-church-lady
@Van Buren:
Well, it certainly does counterpoint the surrealism of the underlying metaphor.
different-church-lady
OK, I’m getting a little lost here: why do we need to have computers generate bad art when so many people are already doing it?
worn
@Enhanced Voting Techniques:
How’s about going on the Tonight Show?