This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
Sora is dead
It turns out that spending hundreds of millions for users to make useless slop videos was having a meaningfully negative financial impact. The bizarre thing is that Disney signed a $1b deal with OpenAI just a few months ago - who fucked up here? Of course, there are many more video AI tools out there, with fewer considerations for copyright law. But for now, Hollywood doesn't have much to worry about, at least on this front.
I tried using Sora for about a month at the end of last year, but I had to stop due to getting banned. Grok Imagine wouldn't ban me, so I've been using that instead. My wild guess is that a social media platform based entirely around AI generated videos like Sora can only exist in a sustainable way if it's explicitly for erotic/pornographic material - there's simply not enough demand for creating or viewing AI generated videos that aren't in that category to get enough users and views to pay for the generations.
I've mixed feelings, since Sora was clearly much better and more flexible than Grok imagine, and so I would've loved to see that develop further, but, at the same time, the lower censorship in Grok and XAI's general attitude towards censorship versus OpenAI's makes me think improvements in Grok is more likely to bear fruit. Of course, without Sora around, XAI has less reason to improve Grok... And Grok Imagine is also still censored, which isn't great, but it's the least worst, at least. In the long run, I'd hope that local video generation will be "good enough," but that'll probably require a world where dual 5090s with 64gb VRAM is considered a quaint little living room computer for sending emails and running old games at a tolerable 25fps, which I'm guessing is within 2 decades.
More options
Context Copy link
I think it's a monetization issue. OpenAI needs to optimize it's revenue per compute. The slop videos take a ton of compute and the users won't spend much on them.
In contrast if they can give Disney a tool with more precise control it's worth a lot to Disney.
Look at a movie like American Sniper. It has a scene where Bradley Cooper is holding an obvious doll because the baby was acting up on set. Afterwards they decided that the expense, delays, an quality issues of fixing the scene with CGI didn't make sense.
If they could have spent $500k in tokens to fix the scene they would have.
Giving directors and editors a tool to tweak things has a lot of value.
More options
Context Copy link
It’s more likely that they couldn’t compete with open source models, Chinese offerings, and other companies specialised in video generation. They also bet on this strange social media app while most AI video generation users want to post their slop on YouTube, Instagram and Tiktok directly. Take a look at the average person’s feed and you’ll see how prevalent AI videos are.
Unscrupulous content creators probably prefer services that let you make videos with little to no censorship, something that OpenAI couldn’t offer.
Which open model is closest in quality?
More options
Context Copy link
More options
Context Copy link
I used Sora 2 for about a week when it came out, then never used it again. I might have used the app for funny video generation, but they insisted on watermarking everything and creating a social media bubble for Sora content, which was always a dumb idea. Meanwhile Sora 1, which actually had a cool discovery feed for generated images, was just sunsetted as well.
The OpenAI ecosystem sucks now in a way it didn’t 3 months ago. Claude is ascendant.
In the most important thing so far - code codex wipes the floor with claude
More options
Context Copy link
More options
Context Copy link
I fucking hate the internet. People refuse to do thier due dillagence, spouting off uniformed one-offs to generate click-revenues without doing thier goddamn due dillage and include some fucking references.
Yes, I'm ranting. Why? Because of one-off remarks on twitter claiming '"Massive investment in AI contributed basically zero to US economic growth last year," per Goldman Sachs'
Clearly, this is a big deal. Quite the claim. Wonderful. Delightful. You'd think they'd have the decency to actually include a fucking link to the article claiming this, but no. More slop for the midwits.
This did not satisfy me. So, I went sluthing, and found the article that sources the claim, which is... a goddamn video interview.
For fuck's sake.
But no. The wizard is undetered. Thankfully, my stubborness in the face of video essays means I have have semi-reliable sources for grabbing youtube transcripts, which I promptly did so to find the money quote. Which I will share for you now;
Interviewer: What surprised you and what didn't in 2025?
Jan Hatzius: Yeah, what surprised us was that the increase in tariffs was bigger than what we had in our baseline assumption, closer to our our risk case, but uh definitely a more pervasive increase in in tariffs.
And that I think explains why US growth was was weaker. The the the shock was bigger. But if I step back from the specific numbers, I think the title of our outlook going into 2025, which was tailwinds, probably Trump tariffs, held up reasonably well. [...] Second driver of growth in 2026 is the fiscal boost. The tax cuts in the one big beautiful bill act which are taking effect mainly in the first half of 2026.
And there's an an aspect here of full expensing of plant and equipment which is going to have a positive impact on the corporate sector and probably on business investment. Tax refunds for households are going to be quite a bit higher than normal. That should support consumer spending. And you know, we're estimating that fiscal policy adds a half to 3/4 of a percentage point to growth in the first half of this year.
And then lastly, financial conditions have been easing as we've gone through most of 2025, at least post liberation day. And that's been for a variety of reasons, but in part because the Fed's been cutting interest rates. And that impulse to growth is also probably going to be most visible in early 2026. So all of that gives us a an above consensus uh growth estimate.
You'll note that one uh one item that may be missing from this list is AI investment. Uh we don't actually view AI investment as strongly growth positive.
I think there's a lot of misreporting actually of the impact that AI investment had on US GDP growth in 2025 and it's much smaller than is often perceived because most equipment most AI equipment is imported and that means there's a positive entry in the investment line but that's offset by a negative entry in the net net exports line And a lot of the AI investment that we're seeing in the US adds to Taiwanese GDP and it adds to Korean GDP but not really that much to US GDP.
Interviewer: Can I just pause there for a moment because at your point and I think this is in the forecast as well is that basically AI investment has been negligible to US GDP growth in 2025. Is that a fair assessment?
JH: Yes. Basically zero. Basically zero.
Interviewer: That seems completely counter to the narrative we read almost every day in financial media. What we hear talked about AI capex and boosting and how that's it's basically supporting the US economy through tariff headwinds and otherwise. Where is the disconnect between the reporting and what you're saying here?
JH: Well, again, I think some people forget that you need to look not just at investment but also at net exports. just from an from a GDP accounting perspective. There's another more technical point which is that some of the AI investment as the AI investment directly in semiconductors isn't actually classified as investment in the national income and product accounts. It's classified as intermediate inputs. So the the national income and product accounts miss that part of investment spending. So we've talked about the the true GDP impact which is still very small but positive positive contributor to growth in in 2025 and then the measured GDP impact which is literally we estimate uh zero.
Interviewer: Okay. I think that's just an incredibly important point to underscore and take forward. And then in 26 you estimate slight increase there as investment comes more online.
JH: We do. We do, but it's still pretty small. I mean, the the the significance of AI and the AI trade and, you know, none none of what I'm saying means that I don't think AI is important, right? It's I think it's very important and it's obviously very important for financial markets, but the specific impulse from AI investment on GDP is still going to be pretty small. Although I do think it's going to be positive to a limited degree this year.
Please excuse any odd errors in the transcript taken straight from youtube. People's direct quotes can sometimes translate oddly to text. Some of the interview has been excised for brevity, as marked by '[...]'
I have suffered through find all of this information, and now you must suffer through reading it. The above discussion can be taken almost straight from the start of the video, should you wish to double-check the above quotes.
And I note all of this above to venture to wonder if we'll see, if not a bubble popping on AI this year, atleast some severe evaporative cooling. It's something to question by this point, but it's always good to recall that markets can remain irrational longer than you can remain solvent.
Hatzius is one of the few economists marked out for his successful predictions and is actually worth listening to; and I agree with his statements about AI.
AI is mostly a solution in search of a problem. At best it may be a somewhat useful assistant when it comes to automating various tasks in the same way critics think of LLM's as little more than a fancy autocomplete but are fairly worthless when measured up against the claims people make about them. At worst the consumer models are like a fucking vagrant living in your basement, with as much as they get shoved deeply into the OS are increasingly pushed into integrations that people are forced to adopt, it feels like your system is harassing and nagging the fuck out of you to use it.
More options
Context Copy link
How can this be?
NVIDIA's revenue for fiscal year 2026 (ending January 25, 2026) was $215.94 billion
Nvidia revenue would be roughly 0.7% of US GDP if it were all in America. Nvidia margin is about 75%. So 25% of revenue goes to manufacturers in Korea or Taiwan. Maybe another 25% is foreign employees, operating expenses abroad. At least 50% of Nvidia should be derived as US economic activity. The chips are designed in the US after all. That's a cool 100 billion dollars or 0.3% of GDP, nothing to sneeze at.
Then there are all the other AI hardware companies like AVGO, the cloud providers like Azure or AWS, the AI companies themselves.
How could people possibly be building these gigantic datacentres and not have that picked up in GDP? https://youtube.com/watch?v=VLgDvjcvURc
If GDP is somehow not measuring the impact of AI investment then so much the worse for the GDP calculators I think. And it's not even clear that this is the case, EY seems to disagree:
https://www.ey.com/en_us/insights/ai/ai-powered-growth
I don't know if EY knows better than Goldman Sachs, I don't have a high opinion of either. However, I think that there are lots of people who want to hear that the bubble is popping and grasp for any sign that it is.
Imagine that it was the 1910s and tractors were the big new thing. Obviously tractors raise agricultural productivity. But maybe they're kind of unreliable, maintenance for this new technology is a bitch, maybe the methods for using them aren't well-established, maybe there's some difficult soil where the tractors get bogged down, maybe fuel distribution in the countryside isn't well-developed. There are lots of conservative farmers around. One could easily produce convincing anti-tractor arguments and examples. But in general, tractors would still remain the future of agriculture, profitable to produce and use. You could derive this from first principles, considering the power of engines and their utility vs horses. There was no tractor bubble, there is no AI bubble.
I'm not so knowledgeable for tractors specifically, but there definitely were car bubbles. When people realized that cars have reached a state close to horse-drawn carriages, will predictably become even better in the near future, and probably replace horses altogether, a host of new companies with various variants of car technology cropped up. Of course, as you should know, almost all modern cars have used the same basic style of engine, the internal combustion engine (and in fact, mostly a specific kind of it).
But it wasn't always so. There was a variety of companies with their own engine designs that overwhelmingly failed. And of course various companies experimented with non-engine related designs that also didn't work out. Some companies successfully made the switch early enough, but many just failed. You could do everything right, correctly predict the dominance of cars, invest in a reasonable company, and lose absolutely everything anyway.
The main difference now is that many of the current competitors are already giants so they can write off a lot of losses without going broke, and it's unclear whether governments will even allow them to outright fail. But a bubble popping on several of them (or even all of them - maybe the real breakthrough will come from a smaller competitor, though I consider it very unlikely) and them losing substantial valuation seems like a foregone conclusion, even if I think that eventually AI will be a technology of the future.
My understanding is that something similar happened for railways, even without the significant differences in engines (?). Revolutionary technology that generated multiple bubbles along the way that bankrupted many people.
More options
Context Copy link
More options
Context Copy link
A technology can end up being dominant and important but still throw off bubbles in the meantime. Stuff like the early Dotcom bubble where everybody could see that the internet was going to be important, and yet funds were allocated to the wrong manifestations of that. Alternatively, some tech is hugely meaningful and yet a motherfucker to actually capitalize on as an individual private company.
But why is this even the right frame of looking at things? If it looks really powerful and useful, probably it is just really valuable. There might be reasons why it isn't but they should be specific.
What has tech been doing for the last 10 years besides AI? I can't think of any great improvement between 2012 and 2022. More irritating ads, having to subscribe to Microsoft Office, VR headsets with a handful of good games... Incremental advancements at best.
But now if I want to do something with a computer, I can get an AI to give me precise instructions or just do it outright. That's a genuine improvement. I don't have to wade through forums or oceans of SEO + ads with a search engine. I don't have to learn to code to make and sell code commercially - they seem to enjoy it and find it very helpful. I can just have it output niche, highly specific pieces of writing just for me, on a whim.
The technology sector has finally contributed something positive after about 10-20 years of resting on their laurels and now people are complaining about a bubble, it seems bizarre to me. This is putting aside all the scientific innovation and prospects of superintelligence.
Cisco had a P/E of 200 back in the day, Nvidia is in the 30s and is the largest company in the world. Where is the bubble? Is it just isolated to the AI pureplay companies like Anthropic or OpenAI who aren't profitable because of all the R&D they're doing?
I know it's cliched to say 'this time it's different' but if it looks different, then it probably is different? Maybe OpenAI has just gone down too many paths and finds that short-form video is not cost-effective since it caters to a population of poor people trying to game porn restrictions and upsets influencers/artists who love to hate AI video. Whereas selling coding AI actually makes lots of money, so they're redirecting compute for business use. In my experience it's far harder to make a profit selling AI to consumers than it is to businesses.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
This may be a canary in the coalmine because it's the apex of structural problems for transformers as a product.
First, the operating costs must have been enormous. Video processing is some of the most costly things you can do on a computer, full spectrum, and it's involved at every layer of processing here. YouTube has famously been a loss leader forever on the storage costs alone, though the usage pattern is vastly different I can't imagine operating the datasets, training and distributing video models is much different.
Second, open weight (or just Chinese) models eat up your moat and don't have the limitations a company is ultimately tied to for PR or legal reasons. In a funny twist or fate, it seems pretty easy to steal back all the knowledge one has stolen by pirating all the movies in the world through "distillation attacks". So innovating on public models can end up just spending money for your competition.
Third, finding product market fit can be tricky even for something as technically impressive as generative AI. It's not clear how this is useful at all for anything but shitty memes and fraud, and hybrid approaches that don't rely on fully generative techniques like DLSS5 could end up being far more practical both to use and to make.
All these are problems every lab is having right now and must solve before the money runs out. That we see people cutting back where they hit hardest may be a sign that at least the money no longer feels infinite.
More options
Context Copy link
It's possible that this is a very specific failure; that arbitrary video generation has no meaningful commercial use cases, and very limited recreational use cases. I used Sora very briefly when it came out, and I found it quite limited, with the substantial cost preventing me from toying-around to figure out what works.
It's also possible that OpenAI is strapped for cash as the AI financial bubble approaches a tipping point. OpenAI's compute committments are based on forcasted exponential revenue growth. If growth is linear or only barely superlinear, then the whole thing falls apart, with everyone holding on as best they can until the next equity sale.
NVIDIA stock is flat over the last 18 months by the way.
Even if they aren't running out of money, they are trying to shape up for an IPO this year. Having a boar-anchor like Sora on the books probably wouldn't look good in an S-1.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link