What is it that these CEOs think they are seeing, that everyone else is missing? I can believe they have inflated egos, but they’re not totally crazy, right? My impression is that Netflix is fairly sober and results oriented, so I’m confused by the whole thing.
> “We should bear in mind that, in general, it is the object of our newspapers rather to create a sensation — to make a point — than to further the cause of truth.”
Bezos bought the Post for clout. Ellison (and his investors) are buying Warner Brothers first and foremost to make money.
Sure. The same goes for most bunker-buster LBOs. Doesn’t mean the sponsors are doing it with the expectation of losing money.
This appears to be Larry Ellison trying to manage succession for his kids. We'll probably see a second major acquisition like this for Megan (who tends to be more progressive leaning) [edit: I was right. Megan Ellison is making moves as well now [0]]
Larry would not be able to do something similar at Oracle today (definitely in the 2000s though), as by the 2010s operational control at Oracle increasingly shifted to operators like Catz, Hurd, and Kurian. It also would have led to bad blood à la the Murdochs.
One heir cultivating the right wing (David) and the other heir cultivating the progressive wing (Megan) buys a level of political impunity for the next generation that is hard to come by.
[0] - https://www.hollywoodreporter.com/movies/movie-news/megan-el...
Capital One to lay off more than 1,100 in latest cuts at Discover Financial HQ - https://news.ycombinator.com/item?id=47270442 - March 2026
To be fair, sometimes they do. Musk saw genuine bloat at Twitter. I’m doubtful one can optimize that much out of WarnerBrothers. But Hollywood isn’t exactly known for being efficient.
That gets tricky: To the extent that an AI is just a tool — along the lines of a trained pair of hands executing a human prompter's specific, detailed instructions — the human prompter might qualify as an "author."
From the U.S. Copyright Office in January 2025:
"The Office affirms that existing principles of copyright law are flexible enough to apply to this new technology, as they have applied to technological innovations in the past. It concludes that the outputs of generative AI can be protected by copyright only where a human author has determined sufficient expressive elements.
"This can include situations where a human-authored work is perceptible in an AI output, or a human makes creative arrangements or modifications of the output, but not the mere provision of prompts.
"The Office confirms that the use of AI to assist in the process of creation or the inclusion of AI-generated material in a larger human-generated work does not bar copyrightability.
"It also finds that the case has not been made for changes to existing law to provide additional protection for AI-generated outputs."
https://www.copyright.gov/newsnet/2025/1060.html (emphasis and extra paragraphing added).
Every single CGI rendered frame of Shrek is protected by copyright because it was human authored. It they used a diffuser to make Shrek 7, the individual frames would not be protected by copyright but their arrangement into a movie could be. That's a hugely different legal situation (for instance, if I chopped it up and made my own remix of it that would be protected).
Short clips are doable, but the coherency duration before it goes weird just isn't there yet, unless you're OK with very short segments (like, single-digit seconds) between each AI-equivalent of a cut. That will probably improve, but the tech isn't there yet for "a lot", it's like CGI in the 90s rather than CGI as it is today.
That's aside from how AI output isn't copyrightable, which may or may not matter depending on if your goal is money or propaganda.
Citations?
I haven't, and would be genuinely curious if it is better than the AI short clip slop that appears on YouTube and really just needs to be taken out back and shot.
I have no doubt that its coming, I just haven't seen it yet.
FWIW, there are models which are better than the majority of the YouTube slop I've seen*; but even then, when I show short clips from e.g. Google's "Flow" generator to others, those others find the results unimpressive. Also, the voice range is weird, my experience playing with them it had a much harder time of creating a decent working-class British accent than Dick Van Dyke.
However, amongst the various issues Flow has, the most obvious is length, because even though Flow lets me "extend" a clip, it's not got enough coherency between each (8 second?) segment when I do.
The second biggest issue is that, as with Stable Diffusion before it, it quite often produces stuff you just don't want. Junk dealers don't care about that and use the first result, an artist can afford to do 10 attempts and pick the right one, and will care to, too. That's something I can at least expect will be solved, or has been on non-free tools (control nets etc.) even though I don't expect slop dealers to ever bother with them.
If each 8 seconds of video was actually Hollywood quality (they're not, in any sense including resolution) and they were coherent when extended (likewise, they're not), as there's 750 times 8 seconds in a 100 minute film, it would still be an existential threat to Hollywood if those segments cost $1333, because that's a million dollars and in the cost range of "micro-budget indie", rather than the tens of millions that even normal-indie films cost (before marketing etc.) or hundreds of millions that blockbusters cost.
I have no idea how video generators scale up in cost with time or resolution, let alone quality, so I don't want to make guesses on how much extra compute (or indeed training data) would be needed for a coherent 2 minute cut-equivalent at 4k. As I don't like the epistemic collapse we're already witnessing with clips from video games being passed off as coverage of warfare, I very much hope this is one of those things which fails to scale up.
* although even with early models, if you wield them as an artist rather than as a slop merchant you get interesting output; you may remember the Harry Potter Balenciaga videos from a few years back, same channel and you can see how things have changed: https://www.youtube.com/@demonflyingfox/videos
This is demonstrating the specific limitation I was talking about. And also, some of those scenes, if you pause on them, they've got the slightly-off vibes of GenAI content, a casual viewer is only missing that because the scenes change so quickly. Something something moonwalking gorilla.
An artist can make a work of art despite the limitations of whatever tool they use (and indeed may get a kick out of doing just that), that doesn't mean the tool can do everything.
I myself am working on a GenAI musical comedy sketch video, but it is only possible with current tech because there's a good opportunity to make cuts every few seconds.
They may become so, I don't know how resolution and temporal coherency scale, but right now Hollywood's biggest problem from China is that China's got 1.5 billion people who are just as capable of learning SFX as anyone in the US.
He arguably single handedly shifted the entire Overton window in the US to the right, and got his preferred candidate elected.
It's reasonable to assume that Ellison is interested in the exact same kind of power, but over the broader culture.
If the argument is that it was a stupid business decision, but he had other motives (clout, etc…), sure whatever.
At the time? Probably. With the benefit of hindsight? Probably not. Twitter would have made a killing on its own licensing its data to AI companies.
I'm reminded of a saying: "Just because they are out to get you, doesn't mean you're not paranoid."
That he was able to cut a lot from Twitter, doesn't mean he cut wisely, for reasons connected with reality. It survived the loss of teams dedicated to ongoing feature development, but the loss of teams dedicated to community management appears (from the outside) to have been the proximal cause of the loss of trust that many advertisers, users, and governments have had with the company. Cutting both at the same time suggests the action was not done from a place of wisdom, but rather that it was luck one of the cuts wasn't critical.
(This is what I expect given a Muntzing strategy: discover which components are load-baring by seeing what happens when they are removed. But when did I learn about Muntzing, and was it after Musk did his thing with Twitter?)