TIL: October 26, 2025 - Weekly Reading: Sora 2 and the AI Bubble

Sora 2 and the AI Bubble(?)

This week I watched several videos discussing Sora 2, OpenAI’s TikTok clone where users create short AI slop videos using real people’s likenesses. The hubris and obvious ethical issues really make me think that there is a kind of “ethics doesn’t matter, we’ll just do whatever”, grifter nihilism at play with tech business leaders.

• SORA: the all Ai TikTok Clone. will slop end creativity? – Casey Neistat

  Casey covers the content creator perspective, and how Sora 2 will likely flood the market with low-effort low-quality AI-generated content. Casey’s key message is that friction in the creation process limited the volume of bad content you would see. More content going into the funnel doesn’t necessarily mean more good content coming out.

  It makes self-expression available to everyone, and I love that. But what happens when you remove the entire process? What happens when all you have to do is like type a couple of words from your bed in a dark room and click a button, and it gives you a piece of video? Then you share it, and then you do that a thousand times a day.

  As good as Casey is, I think he could explain it a bit better. The way I see it, the friction in the creation process forces creators to be creative (who would have thought?) and put effort into the quality of the work. That friction is the heat which forges good content. It’s less about the volume and more about the required effort.

  Aside: Casey is just so good at making engaging videos. From the cream of mushroom soup can representing AI slop, to recording himself buying the funnel that he uses in an explanation later in the video, to his use of AI-slop to provide examples and express ideas. 🤌

• Sora Proves the AI Bubble Is Going to Burst So Hard – Adam Conover

  Adam goes into the ethical issues with Sora 2 as well as the business aspect. Adam’s videos are commentary, but he does a decent job of adding sources.

  He makes some of the economic argument well. OpenAI and others are going to have to make so much money to justify the enormous investments they’re making in data centers and infrastructure, that it’s hard to see how they do that with their current products and business models.

  … in order to pay for all the processing infrastructure they need to build this brave screwed world, Bain estimates AI companies will have to earn $2 trillion annually by 2030. According to the Wall Street Journal, That is more than the combined 2024 revenue of Amazon, Apple, Alphabet, Microsoft, Meta, and NVIDIA combined.

  It’s hard to square the existence of a product like Sora 2 that has a high cost to run, with the high-value things OpenAI has promised it could be doing instead. Like curing diseases, solving climate change, or automating jobs. It makes you think that maybe they aren’t really up to the task so they are grasping at low-hanging fruit (or rotten fruit that’s falling off the tree).

• Give me a single reason why Sora2 should exist. – Hank Green

  Hank’s video isn’t the best produced. It’s basically him ranting, but he makes some good points. He does make similar ethical points. All three videos touch on the use of real people, the lack of consent, and the sheer ease of creating content that is misleading or slanderous.

  He points out an interesting thought that previously the argument was that services weren’t responsible for user-generated content. But with AI the platform actually has a hand in creating the content so do they also bear some responsibility for it?

  And if you trot out section 230 of the Communications Decency Act and say we didn’t do that, the person put in the prompt. It’s not OpenAI’s fault. We can’t be held responsible for the content of our platform even though our computers enabled its creation and also did the actual creation of turning a single sentence into that monstrosity.

• The State of the AI Industry is Freaking Me Out – Hank Green

  Hank made a second video that is more general about the AI industry and how it’s funding itself. It’s not terribly well researched but it does note a pattern in how companies are investing and lending to each other in a kind of closed ecosystem that seems unsustainable. The whole thing seems propped up by HUGE amounts of investment money which is chasing a return that has a worryingly high chance of never materializing.

  He points out is hard to believe that OpenAI selling its end product, making $12 billion a year, is supposed to sustain hundreds of billions of dollars of investment in infrastructure going to NVIDIA and data center construction.

  … the chip isn’t the product. The AI service to consumers is the product. And right now, that is an order of magnitude under the amount of money that is being spent just on hardware. Not even including staff time and the power and the water and the buildings and all of that stuff. The amount of money being made by AI companies is an order of magnitude lower than just the money that is going to NVIDIA every year.

  Hank responded to some comments on this video in a follow-up video where he clarifies a few points. Notably, he points out that he thinks AI is a transformative technology. It just won’t generate the kind of returns that justify current investment levels. He also seems to think this strongly enough that he’s changed his investment strategy to diversify away from the S&P 500 because of its heavy exposure to AI-related stocks.

• Bubble, Bubble, Toil and Trouble – Zvi Mowshowitz

  This article provides a bit of counterbalance to the AI bubble doom and gloom videos above. Zvi finds the bubble arguments weak and unconvincing. This is partly due to the fact that the Nasdaq forward P/E ratio is something like 28x compared to 70x during the dot-com bubble and the track record of growth with AI companies is strong so far.

  I think the danger is that the economy is a bit weak aside from the AI investments, and the expectations are just so high that it’s hard to believe they will sustain the expected growth given the relative costs.