LightNews LightNews
wettyds.bsky.social
@wettyds.bsky.social
85 followers 160 following 66 posts
Posts Replies Media Videos
Reposted
moreperfectunion.bsky.social
OpenAI pirated large numbers of books and used them to train models.

OpenAI then deleted the dataset with the pirated books, and employees sent each other messages about doing so.

A lawsuit could now force the company to pay $150,000 per book, adding up to billions in damages.
November 4, 2025 at 3:30 PM
Reposted
theverge.com
Studio Ghibli, Bandai Namco, Square Enix demand OpenAI stop using their content to train AI
Studio Ghibli, Bandai Namco, Square Enix demand OpenAI stop using their content to train AI
No more Spirited Away and Demon Slayer in Sora 2.
buff.ly
November 3, 2025 at 5:50 PM
Reposted
edzitron.com
Down 11.5% today, honestly respect to Mark Zuckerberg, late to the party, spending way too much, pissing off everybody, doing nothing but waste money, and then seeing his stock wilt as he has no way to prove it's worth it. I hope he is the first to pull capex lol
cnbc.com CNBC @cnbc.com · 11d
Meta shares are plunging in heavy trading, on pace for their worst day in three years. Here's why: cnb.cx/4hAPrzH #CNBCData
October 30, 2025 at 3:42 PM
Reposted
jtp.bsky.social
October 29, 2025 at 1:42 AM
Reposted
edzitron.com
betamax is back - as a laser disc
October 25, 2025 at 3:53 AM
Reposted
oweeeeendennis.bsky.social
It's crazy how much workers do to try and save these studios from themselves.
culturecrave.co Culture Crave 🍿 @culturecrave.co · 17d
WGA plans to help block Paramount from buying Warner Bros

"Combining [them] would be a disaster”
October 24, 2025 at 4:18 PM
Reposted
oweeeeendennis.bsky.social
October 24, 2025 at 4:34 PM
Reposted
genc.newgrounds.com
21st Century Schizoid Imp
[ twitch.tv/genczilla ]
October 21, 2025 at 2:10 AM
Reposted
culturecrave.co
“Fuck AI”

— Guillermo del Toro

🎥 Vanity Fair
October 18, 2025 at 3:20 PM
Reposted
edzitron.com
Sora is a paper machete. It is low-quality, impossible to make consistent or reliable, while also being incredibly expensive to run. Every single “good” video you’ve seen is cherrypicked from dozens of terrible ones to make you feel scared that this can create entire movies’ worth of content.
hareandthere.bsky.social Dustin @hareandthere.bsky.social · 25d
Curious to hear your perspective on Sora. It seems to be the only real (and terrifying) AI product to hit the market lately. How much mileage are they going to get out of this thing that needs to be regulated out of existence?
October 16, 2025 at 12:59 AM
Reposted
chibdm.bsky.social
Robin Williams' daughter has some quality thoughts on AI slop
October 6, 2025 at 8:48 PM
Reposted
cynthia.dev
September 25, 2025 at 1:26 PM
Reposted
goodbyellow.bsky.social
KPop Demon Hunters in Fortnite?👺😈
October 3, 2025 at 7:15 PM
Reposted
edzitron.com
Sora 2 is an act of desperation - an ugly, bloated app that deliberately violates copyright and exists ONLY as a marketing tool.

Do not fear Sora. It's inconsistent, watermarked, most outputs are garbage, and OpenAI *cannot afford* to run it like this long-term.
www.wheresyoured.at/sora2-openai...
A sidenote for those scared of Sora 2: I also want to speak from the heart here to anyone worried about generative video, because the reality of Sora 2 is far funnier and more pathetic than OpenAI would have you think. The best way to look at Sora 2 is that it’s a giant, expensive slot machine While Twitter is full of “impressive” looking clips, Sora 2 is still a generative model, and the majority of the clips it produces are ugly, weird, unsettling and full of weird glitches and animations. While it’s superficially impressive that it can do anything close to real video footage, the practical reality of Sora 2 is that it isn’t particularly practical. Getting the output you want once is challenging enough before you realize that there’s no way to guarantee that anything stays consistent between shots, as you can’t edit any of these outputs beyond editing the prompt that created them. In my testing, Sora 2 continually changed how it would interpret simple things like “anime style,” and the only way to make it do consistent outputs was to literally instruct it to make something in the style of a copyrighted material like the anime Demon Slayer. The Sora 2 app also watermarks each video, whether you use the app or Sora.com to generate the video. It’s also onerously expensive for OpenAI to run. I understand the “potential for deepfakes” and “potential for replacing humans” feels real here, but the actual reality is that you cannot guarantee Sora will do anything right, and even with precise prompting it will sometimes decide to do something else entirely. While some of the videos might look realistic, making those realistic videos into anything precise is difficult-to-impossible. Sora 2 is a mirage, one built to scare you into believing that OpenAI has something that will last the test of time, versus the result of a company that loses $2.60 for every dollar it makes, and has just created a gigantic burden on its infrastructure that will only increase those lo… A sidenote for those scared of Sora 2: I also want to speak from the heart here to anyone worried about generative video, because the reality of Sora 2 is far funnier and more pathetic than OpenAI would have you think. The best way to look at Sora 2 is that it’s a giant, expensive slot machine While Twitter is full of “impressive” looking clips, Sora 2 is still a generative model, and the majority of the clips it produces are ugly, weird, unsettling and full of weird glitches and animations. While it’s superficially impressive that it can do anything close to real video footage, the practical reality of Sora 2 is that it isn’t particularly practical. Getting the output you want once is challenging enough before you realize that there’s no way to guarantee that anything stays consistent between shots, as you can’t edit any of these outputs beyond editing the prompt that created them. In my testing, Sora 2 continually changed how it would interpret simple things like “anime style,” and the only way to make it do consistent outputs was to literally instruct it to make something in the style of a copyrighted material like the anime Demon Slayer. The Sora 2 app also watermarks each video, whether you use the app or Sora.com to generate the video. It’s also onerously expensive for OpenAI to run. I understand the “potential for deepfakes” and “potential for replacing humans” feels real here, but the actual reality is that you cannot guarantee Sora will do anything right, and even with precise prompting it will sometimes decide to do something else entirely. While some of the videos might look realistic, making those realistic videos into anything precise is difficult-to-impossible. Sora 2 is a mirage, one built to scare you into believing that OpenAI has something that will last the test of time, versus the result of a company that loses $2.60 for every dollar it makes, and has just created a gigantic burden on its infrastructure that will only increase those lo… Material, And Allowed Users To Generate Video Using It As A Means of Going Viral OpenAI has clearly ingested hundreds of hours of copyrighted material as a means of making Sora 2 go viral, and launched the product knowing that users would seek to produce copyrighted works. OpenAI is likely gambling that it can prevail in court, or that rightsholders will somehow back off, or that they can be paid off, but in doing so it is playing with fire. Based on my tests, it’s clear that OpenAI has ingested popular anime titles like Demon Slayer and Chainsaw man, hundreds of hours of gameplay of Zelda: Breath of the Wild, and multiple different Grand Theft Auto games, and what I’d guess is hundreds more hours of footage of real people from YouTube, all in the belief that nobody would dare sue or that it could tie them up in months or years of court cases. I cannot express how silly that is. Japan’s major publishing houses are extremely litigious, and successfully sued multiple online piracy sites (and won $14 million from one earlier in the year) for illegally hosting works from manga like One Piece and Kingdom. Shueisha, who owns Shonen Jump (which publishes manga and anime like Naruto, One Piece, and Demon Slayer) has such sway that somebody pretending to be the company was able to issue a number of different copyright infringement claims across social media, taking down everything from images from manga pages to fan art and fan fiction. Interestingly, it appears they’ve trained on hours of audio too. In one generation involving Naruto, it generated a direct copy of the voice of Naruto voice actor Maile Flanagan. It’s too early to tell what the legal action will be — or if rightsholders will simply demand OpenAI stop generating their copyrighted characters — but OpenAI’s flagrant violation of Shueisha’s material will likely cause issues for its joint venture with SoftBank, SB OpenAI Japan, which Reuters reported a few weeks ago is “significantly behind schedule.”
October 3, 2025 at 4:48 PM
Reposted
goobdubious.bsky.social
September 10, 2025 at 6:29 AM
Reposted
glitcharttv.bsky.social
Me adding every extension to kill AI garbage

Life is good.
September 25, 2025 at 12:43 AM
Reposted
mikesenters.bsky.social
This is the line that Dems need to push. It can't be simply. "Political violence is bad" it MUST be "Political violence is bad and Trump and his ilk are responsible for creating conditions ripe for political violence"
heartlandsignal.bsky.social Heartland Signal @heartlandsignal.bsky.social · Sep 10
Illinois Gov. J.B. Pritzker (D) on the shooting of Charlie Kirk:

"Political violence, unfortunately, has ramped up in this country ... I think there are people who are fomenting it in this country. I think the president's rhetoric often foments it."
September 10, 2025 at 7:56 PM
Reposted
edzitron.com
Oh yeah we're headed toward the end of this bubble alright, HELL yes. A Chamath SPAC is the sign that there's too much dumb money. A magreaux dog for the markets, worse than Cramer
techmeme.com Techmeme @techmeme.com · Aug 18
American Exceptionalism Acquisition, a SPAC led by Chamath Palihapitiya, files for a US IPO, seeking to raise $250M (Kia Kokalitcheva/PitchBook)

Main Link | Techmeme Permalink
August 19, 2025 at 3:01 AM
Reposted
edzitron.com
I genuinely think GPT-5 may bring down generative AI. It broke the illusion with normies who thought ChatGPT was a friend or confidant or "brain storm machine," and I believe its new architecture is doing the very opposite of making it cheaper or more efficient.
August 14, 2025 at 1:04 AM
Reposted
edzitron.com
GPT-5 was meant to be OpenAI's iPhone moment. Instead, it felt like the last Valentine's Day before a breakup. This underwhelming launch feels like the end of the GPT era, and the last trick in Clammy Sam Altman's arsenal.
www.wheresyoured.at/the-enshittification-of-generative-ai/
The Melancholy of OpenAI I’m struggling to find articles about it, but I swear, for the last few years, people have been acting like GPT-5 would be some sort of industry-changing event, tantamount to the original iPhone. Instead, GPT-5 feels like the last Valentine’s Day before a breakup. I’m serious, there’s a genuine sadness to everything going on in AI — a kind of pained, feigned excitement for a new model that doesn’t really sound like a new model, and even Simon Willison, a well-intentioned, well-versed LLM-knower, can only muster the limp praise of “verdict: it’s just good at stuff.” How utterly melancholy! How deeply sad! We sit at what amounts to the end of, at the very least, the end of the GPT era, of the period of S development in generative AI when there were things to look forward to and when there was hope that OpenAI would create something impressive. This was, in many ways, the last real reason for anyone to take OpenAI seriously. After this, what exactly is there to wait for from this company? Its next model that becomes the top of some sort of obscure LLM benchmark for a week or two before Google or Anthropic does nearly the exact same thing? What, exactly, are we building all these data centers for? Why are we doing this? To what end? What necessitates the environmentally and societally destructive sprawl of generative AI? Not that I ever believed that OpenAI would actually impress me, but is this really the best that Sam Altman has to offer? Is this why OpenAI has incinerated ten billion dollars since June? Is this what necessitated raising another $8.3 billion, with plans to raise another $22.5 billion to follow? Why are we acting as if there is not, on some level, a cashout happening at OpenAI right now? It’s trying to set up an employee share sale at a valuation that would make it worth more than Netflix at $500 billion. What, exactly, is it about OpenAI that makes this company worth $300 billion (Salesforce, as of writing this sentence, has a market capitalization of $243 billion), let alone $500 billion? What about GPT-5 justifies this company’s existence? Are these tools truly innovative? Are they truly useful? How is this company making a billion dollars a month, and what is it about its product that makes 20 million people pay $20-a-month for it? Are any of the 5 million business customers the ones that OpenAI has offered a month of ChatGPT Team to for a dollar? What’s it’s churn? How much money does OpenAI actually make? Investors plowing more money into OpenAI are not simply wasting their money, but enabling the reckless and erratic sprawl of Sam Altman. Nothing about this company is innovative, there is nothing impressive happening, and it is shameful that this has been allowed to go this far. This company is set to burn at least $8 billion this year according to The Information — but I think it’s going to burn far more, based on the fact it raised $10 billion in June, then $8.3 billion in August, and plans to raise another $20 billion or more by the end of the year. What’s next? Will Anthropic end up doing exactly the same thing, then also raise a lot of money, then also have an employee share sale, then also release the newest, most powerful model of all time? I can’t wait! Really though, does anybody feel excited about what’s happening? Is anybody having fun? Is anybody truly hopeful for the future based on the launch of GPT-5? Or are they just going through the motions until somebody finally cracks and admits that things aren’t going to work out?
August 8, 2025 at 5:26 PM
Reposted
radiodeadair.com
If you want to know what's really going on with "AI" filmmaking, here's a demonstration of some of the most expensive technology available by a professional.

The output is near constant garbage.
Google Veo fails week 4: the final faildown
When you’ve read this post, go watch the worked examples on YouTube. [YouTube] Our intrepid friend Aron Peterson, a.k.a. Shokunin Studio, is a media production guy. He follows the AI image generato…
pivot-to-ai.com
August 8, 2025 at 12:50 AM
Reposted
litbowl.bsky.social
GENERATIVE AI IS THEFT. And it is now officially in the legal record as precedent, argued powerfully and beautifully and eloquently and forcefully by Judge Chhabria.
kint.bsky.social Jason Kint @kint.bsky.social · Jun 25
And...yes. Another big AI decision. When a win is maybe not a win. The AI training is fair use cult may be very disappointed by Meta's Summary Judgment "victory" in its copyright suit. Be very very careful not to be spun by Meta's proxies and advocates.
READ THE ACTUAL OPINION CLOSELY. /1
June 25, 2025 at 11:31 PM
Reposted
kint.bsky.social
On a final note, the Court also notes this decision only applies to the thirteen book authors. The rest of the class can sue again on the same claims. And mark my words they will. /6
June 25, 2025 at 10:59 PM
Reposted
indagare.bsky.social
When Galactus is attacking Earth and Dr. Doom is the one keeping Galatus at bay, one roots for Dr. Doom.

This doesn't make Dr. Doom a moral person or someone one would support in almost any other situation.
June 11, 2025 at 6:06 PM
Reposted
edzitron.com
If I had to guess this is them looking to settle this whole debate once and for all in a way that sets a legal precedent that makes the entire image generation paradigm unviable to impossible
June 11, 2025 at 4:38 PM