Sora is one of the fastest growing apps in history.
When they add ads to ChatGPT, they'll make bank. I use ChatGPT way more than I use Google now.
It doesn't give good recommendations because the training set is so out of date, and I find it unusual that that anyone would use it for product recommendation.
Lastly, "fastest growing" in the short term does not amount to much in the long term.
I would say the exact same for Google.
What isn't SEO'd to hell is crowded out by half a page of paid search ads.
I'm not a typical consumer though. I don't think I've ever bought anything via an ad.
The only search ads I click are the ones where Google FOMO-extorted the brand into buying ads for their own trademark. (That shit ought to be illegal given the monopoly levels of search capture Google has. Everyone having to buy up search clicks for their own brands and company names is comical. A protection racket.)
If I would have asked an LLM, it would have told me to buy an all-metal hotend that costs $70 or more, based on outdated advice.
I don't trust LLMs for out of the box thinking because they have no idea what is going on in the real world. They are fine for discussing general aspects of well-documented and often-discussed things like taxes, cooking, or gardening.
Here's something I implore readers think about, just to help ground your reality in the numbers we're talking: The super bowl gets ~120M viewers. During the four hour event, this year, they turned ~$800M in advertising revenue. What you're thinking is "I see where you're going with this, but that's higher value advertising, its not the same on..." and I'll stop you right there and ask: How much revenue do you think Meta makes every day? The answer: ~$450M. Not far off.
OpenAI will make so much money if they figure out advertising. Incomprehensible, eye-watering amounts of money. They will spend all of it and more building data centers. If your mental model on these companies is still based in some religious belief that they need to achieve AGI to be profitable, and AGI isn't possible/feasible: you need to update your model. AGI doesn't matter. People want to stop thinking and sext their chatgpt robot therapist. You can hate it, but get with the program and stop crystalizing that legitimate hate for what the future looks like as some weird fantasy that OpenAI is unsuccessful and going to implode.
The fear of an AI bubble isn't that AI companies will fail, it's that a downturn in the AI "bubble" will lay bare the underlying bear market that their growth is occluding. What happens then? Nobody knows. Probably nothing good. Personally I think much of the stock market growth in the last few years that seems disconnected with previous trends (see parallel growth betwen gold and equities) is based on retail volume and unorthodox retail patterns (Robinhood, WSB, et. al) that I think conventional historical market analysis is completely unprepared for. At this point everything may go to the moon forever, or it may collapse completely. Either way, we live in a time of little precedence.
It isn't? What is stopping companies from building on GPT-OSS or other local models for cheaper? The AI services have no moat.
I agree with your second paragraph. The boom in the AI market is occluding a general bear market.
Right now there is an efficiency/hardware moat. That's why the Stargate in Abilene and corresponding build outs in Louisiana and elsewhere are some of the most intense capex projects from the private sector ever. Hardware and electric production is the name of the game right now. This Odd Lots podcast is really fresh and relevant to this conversation: https://www.youtube.com/watch?v=xsqn2XJDcwM
Local models, local agents, local everything and the commodification of LLMs, at least for software eng is inevitable IMO, but there is a lot of tooling that hasn't yet been built for that commodified experience yet. For companies rapidly looking to pivot to AI force multiplication, the superscalers are the answer for now. I think it's a highly inefficient approach for technical orgs, but time will create efficiency. For your joe on the street feeding data into an LLM, then I don't think any of those orgs (think your local city hall, or state DMV) are going to run local models. So there is a captured market to some degree for the the current superscalars.