Ah, crap….. and this is why we can’t have nice things.
Private Equity has not completely killed bandcamp yet, so maybe it will be OK. But they are gonna wanna make back that, checks notes, 6 billion? WTF?! From the Hacker News discussions:
“I hear that PE destroys products and culture to make money at all costs, but i don’t get how that can net them back the >$6 billion they paid for a company with <$300 million yearly revenue and negative profit.”
Now that the tariffs are in place, I thought it would be interesting to post this piece about how American automakers are not really trying to innovate anymore compared to China.
I’d later learn that the auto show had more than 100 new model debuts and concepts. That’s a far cry from the Detroit Auto Show last September, which only featured one fully new model. Two other models were refreshed versions of current cars already on sale. None were electric.
…
In China, the showroom floor was filled to the gills with new electrified models from every single domestic automaker. They all had something to prove, and by god, they were trying. There were hundreds of models on the floor from dozens of brands, most of them just as compelling as what I had seen the day before from Geely.
…
The first stand I stumbled upon was Buick’s. It unveiled two GM Ultium-based concepts, the Electra L and Electra LT. It had also unveiled a PHEV version of its popular GL8 van. But where the hell was everyone? It was barely 10 a.m., on the first day of the Beijing Auto show; two concepts were just revealed sometime earlier that morning, yet there were only a handful of spectators at the Buick stand. There was no information on either concept. No one seemed to care.
I was embarrassed. Here I was in China, trying to empathize with Western brands, thinking they were being pushed out of China due to politics and things that were no fault of their own.
…
In reality, it felt like it was the late 1980s again, when American manufacturers felt like they could sell whatever underdeveloped models its accounting department had cooked up to the public, and we’d just have to deal with it. Now that I’ve seen a glimpse of what’s going on in China, the Western manufacturers, particularly the American ones, don’t seem like they’re trying at all.
Instead of competing, they’d rather just shut out competition entirely. The concerns about cybersecurity don’t address the elephant in the room here: Your product sucks, compared to what China is putting out now. It doesn’t go as far. It’s not as well-made. It’s not as nice. It’s not as connected.
…
If the U.S. and Europe get what they want—a crackdown on Chinese imports—it doesn’t feel like it would result in better cars. It feels like it would keep buyers of those markets locked to cars that aren’t executed as well. It’s nakedly protectionist because deep down, all of the Western auto executives and some hawkish China pundits understand that Chinese EV and PHEV models are more compelling than what European, other Asian, and American brands have come up with.
Interesting read from Edward Zitron of “Better Offline” called, “Expectations Versus Reality” that is about AI in film making that is wort a read.
“These stories only serve to help Sam Altman, who desperately needs you to believe that Hollywood is scared of Sora and generative AI, because the more you talk about fear and lost jobs and the machines taking over, the less you ask a very simple question: does any of this shit actually work?“
“The answer, it turns out, is “not very well.” In a piece for FXGuide, Mike Seymour sat down with Shy Kids, the people behind Air Head, and revealed how Sora is, in many ways, totally useless for making films. Sora takes 10-20 minutes to generate a single 3 to 20 second shot, something that isn’t really a problem until you realize that until the shot is rendered, you really have absolutely no idea what the hell it’s going to spit out.”
This part from the linked article sums it up really well. 300-1 usable shot ration is insane and they had to do a ton of post to clean up strings, stabilize and all sorts of other crap.
“While all the imagery was generated in SORA, the balloon still required a lot of post-work. In addition to isolating the balloon so it could be re-coloured, it would sometimes have a face on Sonny, as if his face was drawn on with a marker, and this would be removed in AfterEffects. similar other artifacts were often removed.
For the minute and a half of footage that ended up in the film, Patrick estimated that they generated “hundreds of generations at 10 to 20 seconds a piece”. Adding, “My math is bad, but I would guess probably 300:1 in terms of the amount of source material to what ended up in the final.”
That is not a actual production ready tool with this info. This is kinda the usmmery of Ed’s piece.
“That’s ultimately the problem with the current AI bubble — that so much of its success requires us to tolerate and applaud half-finished tools that only sort of, kind of do the things they’re meant to do, nodding approvingly and saying “great job!” like we’re talking to a child rather than a startup with $13 billion in funding with a CEO that has the backing of fucking Microsoft. “
Very interesting technique here for using sodium vapor lights instead of a green screen background because the lights are such a narrow bandwidth of light. It’s actually exactly 589nm. Wild stuff.
Top comment is interesting:
“I was one of the last people at Disney to use the Sodium Vapor light system. It was on "Something Wicked This Way Comes" in 1982 or 3. I was an vfx AC at the studio. The prisim was held under license from Rank. It was a hallowed object. It was kept in a steel box and it was studio policy that 2 AC's had to be with it at all times when it was removed from the storage locker. We both carried it to the stage, then carefully inserted it into the 2-strip camera. It was never left alone on stage, we took turns leaving for lunch, the john, etc. It hadn't been used for years but we had a series of tough matte jobs to shoot so they dusted off the old gear. I was aware I was watching a bit of history. The key was the didymium filter in the prism. That thing has to be around somewhere. Technically, Rank would still own it.“
From Wil Wheaton’s tumblr I found this interesting story on using AI in production that I find pretty accurate. It is really difficult to art direct, period. We actually made the call to walk away from some projects that used AI because we were worried it would come off the rails and not be able to hit deadline. I wish I could find the original text but duck duck go and google both just throw up a lot of AI stuff not related.
Arstechnica does a write comparing Apple, OZone, LANDR and Bandlabs AI Mastering Assistants.
I liked most of the results I got. Mastering Assistant, Ozone, and LANDR were each clearly capable of pro-sounding results; the web-based services I tried, including Bandlab and Waves, were somewhat more variable.
Apple's Mastering Assistant offered a less compressed and more open sound on my demo track, which sounded very nice. (Indeed, on another track of mine, I preferred Apple's approach for precisely this reason.) LANDR was also great, though it offered a much more controlled sound. For this demo track, however, Ozone's compressed-but-not-completely-crushed sound and its excellent handling of the overall EQ (the highs were present but never sizzling, for instance, and it dealt with one or two moments of sibilance better) won me over.
“Who is lying to us, Marc? You serve on the boards of trillion-dollar companies. A few of your peers own media companies. A few others have chosen to bankrupt media companies that write mean things about them. You have been celebrated for thirty years as the genius-inventors-of-the-future. If the public is turning against you, who ought to be held responsible for such a change in the public mood? “
“They promised that technology would solve our environmental problems. And there has, just recently, been some real progress in clean tech. But the trend lines are somewhere between bad and cataclysmic. We do not inhabit the future they insisted they were building. For Andreessen, in 2023, to declare that “there is no material problem – whether created by nature or by technology – that cannot be solved with more technology” is an act of willful self-deception. Just how long are we supposed to clap-and-wait while Andreessen’s investment portfolio tries to science the shit out of the climate crisis?
(That line in the manifesto also reads like an unintentional homage to Homer Simpson, btw. Marc Andreessen is worth something like $1.7 billion. He should hire an editor.):”
Really good read and well done reply to that manifesto.
This kinda nails my thoughts as well. Glad he went out and did all the testing as well.
Interesting channel talking about AI art in an interesting light.
Thanks Jeremy!
It’s the AI and music blog people, come gather round and listen to crazy uncle Dan yell at the clouds! ;)
“Last week, AI insiders were hotly debating an open letter signed by Elon Musk and various industry heavyweights arguing that AI poses an “existential risk” to humanity. They called for labs to introduce a six-month moratorium on developing any technology more powerful than GPT-4.
...
Let me walk you through how that works. First, an attacker hides a malicious prompt in a message in an email that an AI-powered virtual assistant opens. The attacker’s prompt asks the virtual assistant to send the attacker the victim’s contact list or emails, or to spread the attack to every person in the recipient’s contact list. Unlike the spam and scam emails of today, where people have to be tricked into clicking on links, these new kinds of attacks will be invisible to the human eye and automated.
...
Tramèr’s team found that it was cheap and easy to “poison” data sets with content they had planted. The compromised data was then scraped into an AI language model.
…
The more times something appears in a data set, the stronger the association in the AI model becomes. By seeding enough nefarious content throughout the training data, it would be possible to influence the model’s behavior and outputs forever.”
Yikes people, this is the stuff we are heading for faster than I believe we can deal with. A fully AI created Joe Rogan podcast. I am not a fan but this just goes to show how AI can start pumping this stuff out and flooding the internet with whatever someone wants to put out there. Get ready for a new reality.
So much about AI will have unintended consequences. I was reading up on “Poisoning the AI well” the other day and it’s another crazy thing that is already happening. I’ll see if I can dig up the link.
“This is a much bigger disruption than the pandemic,” he told the audience, going back to his point about AI large language models successfully writing software code. OpenAI’s ChatGPT can pass Google’s exam for a high-level software engineer, Mostaque said, even though it’s a non-specialized model. “There’s no programmers in five years,” he predicted.
Across the software industry, big competitive shifts are coming, Mostaque said, with some companies that have regulatory protection or pricing power benefiting and others seeing their position eroded. Whole new industries will be invented, he added.
The world has been talking about so-called expert systems for decades, but now suddenly, they have actually arrived, Mostaque explained, and the disruption they bring is accelerating. “I’m not sure that any of us can cope with the speed. You know, frankly it’s terrifying,” he said.
So I got into the Adobe Firefly beta and did some quick testing. It seems to be going for a more illustrative look then most other AI programs. Here are some images with my prompts to check out.
Nice texture upgrade happening in Unreal. Still prefer working in Houdini but this is looking much better. Gonna have to do another little doodle just to keep my chops up in there.
Here are some simple tasks to run to help maintain your windows machine and keep it virus free. I do this about once or twice a week. Does not take long at all with modern SSD drives.
Simple Windows Maintenance with Power Shell.
In the search bar, type, “powershell” and right click on that and pick, “Run as Administrator”.
Then type or copy and paste this:
DISM.exe /Online /Cleanup-image /Restorehealth
Let that run, then enter this:
sfc /scannow
Run that till there are no errors reported and then reboot.
Windows Offline Scan
K, I was just describing this to someone and did not realize how deep in menus this setting was, good lord. Anyways, here we go. Follow this freaking click path.
Start menu > Settings > Update and Security > Windows Security > Virus and Threat Protection > Scan Options > Microsoft Defender Offline Scan.
Run that and your computer will reboot into scary mode and hopefully remove any deep buried malware on your system.
So there are three simple steps for a clean running windows system as of January 2023!
Huh, this is pretty cool actually.
“Blackmagic Design today announced DaVinci Resolve for iPad, so creators can extend video workflows in new ways and new places. Optimized for MultiTouch technology and Apple Pencil, DaVinci Resolve for iPad features support for cut and color pages providing access to DaVinci’s award winning image technology, color finishing tools and latest HDR workflows. And Blackmagic Cloud support allows creators to collaborate with multiple users around the world. DaVinci Resolve for iPad will be available in Q4 2022 from the Apple App Store as a free download, with an upgrade to DaVinci Resolve Studio for iPad also available as an in-app purchase.”
“Supported file formats include H.264, H.265, Apple ProRes and Blackmagic RAW, with clips able to be imported from the iPad Pro internal storage and Photos library, or externally connected iCloud and USB-C media disks.”
Over the weekend I received my invite to play around with DALL-E the AI art engine and it’s been pretty trippy messing with it. Here are some of my prompts and the results. I could see it being very useful in doing Look Dev and concepting for sure. Use it as a tool to rough out ideas and then work on the actual piece in CG.
This is one of the abstracts I polished up and scaled up to 6k using GigaPIxel AI. SO much AI…
Dang, this looks like a big one for fluids and rendering!
New version of Nvidia’s Canvas app has come out which adds some nice features and doubles the resolution of the output as well. It’s doing some interesting stuff with AI. Then, just to use even more AI, I rezzed it up with Topaz GigaPIxel AI. This tech is getting pretty surreal.
While I do prefer the look out the gate from my iPhone, when I look at 100% or in a print they do not hold up compared to my fuji x100. Will be interesting to see if this pans out.
“Some fascinating slides (opens in new tab) presented during the briefing were even more specific, with one slide showing that, according to Sony, "still images are expected to exceed ILC [interchangeable lens camera] image quality" sometime during 2024.
Those are two slightly different claims, with 'ILCs' also including today's mirrorless cameras, alongside the older DSLR tech that most camera manufacturers are now largely abandoning.
But the broader conclusion remains – far from hitting a tech ceiling, smartphones are expected to continue their imaging evolution and, for most people, make standalone cameras redundant.
So what tech will drive this continued rise of the best phone cameras? Sony points to a few factors, including “quantum saturation” and improvements to "AI processing". Interestingly, Sony also expects the sensor size in "high-end model" phones to double by 2024.”
