The one possible hope here is that since these things started as paid services, we know subscriptions are a viable and profitable model. So there's a market force to provide the product users actually want, which does not include ads.

If OpenAI or the other players are pushed toward expanding to ads because their valuation is too high, smaller players, or open source solutions, can fill the gap, providing untainted LLMs.

Why wouldn't a company monetize both ways? Paid video streaming services still show ads, and when I pay for a movie in theaters, they're still doing product placements.

It makes the service worse. I won't pay for a streamer that uses ads. If they start doing that I'm out. Ofc, that doesn't mean it's not a net win for them across all customers, but it does mean there are a subset of customers who are now willing to pay for a different service, and the market has an incentive to service that.

I really hope LLM becomes a commodity at some point, so self hosting becomes easy.

Is there any reason to believe the current subscription models are viable and profitable outside of huge burn rates? Uber rides are now 4-5x what they cost when they were starting up, but uBer was disrupting an entrenched market with extremely high prices. Even today Uber's are still typically less than 50% of what a Taxi previously cost.

If LLMs are disrupting search then they would have to adopt a similar monetization strategy to be profitable. The major issue with that is LLMs are many orders of magnitude more expensive to run that a search engine.

Citation for subscriptions as a profitable model? Revenue may be high, but actual profit is far into the negative at this point I thought.

Netflix seems alright.

Sorry, I thought we were talking about products like OpenAI.

Obviously subscriptions work for some products that have lower operational costs, but I don't believe that to be universally true for AI as a service.

Look at Prime. They will do both. Paid service plus ads. And ad-influenced LLM output will be hard to recognize.

I agree that will probably happen, but I don't think it's a realistic way to exploit information asymmetry like the article describes. I can't imagine a sleazy car salesman or plumber being able to accurately target only the guy they're trying to rip off right now with expensive targeted advertising like that

Who's economy? Yours?

Because once I have an intelligence that can actively learn and improve, I will out-iterate the market as will anyone with that capability until there is no more resource dependency. The market collapses inward; try again.

> Because once I have an intelligence that can actively learn and improve

Great news - you already do.

*whose

Google is definitely doing it. I was searching one term that later turned out to be an euphemism for suicide and what I got was something about wooden flooring made by this and that company.

Yeah but... running an LLM is braindead simple now with Ollama, someone with a little bit of knowledge could run their own or spin up an LLM backed service for others to use.

It isn't like Google search where the moat is impossibly huge, it is tiny, and if someones service gets caught injecting shit like that into prompts people can jump ship with almost no impact.

LLMs without a search engine attached suck for product reviews.

Yes but what happens when you don't need to even buy "products" anymore because you have a 3d printer at home and you just need schematics?

I assume that you're going to 3d print the mines that you use to build the oil rigs feed the chemical plants that you use for producing the filament, right?

haha. i'm imagining the luxurious comfort of a solid 3D-printed t-shirt. i'll never want for the retail experience again!

I know this is in jest but do you need tshirt reviews?

not online, but i am partial to something a cut above a standard Hanes or Gildan, at least for workwear.

I’m honestly pretty paranoid that this is already happening - I treat specific product recommendations from LLMs the same way I treat ones I sit up on Reddit - they could so easily simply be paid advertisements, smuggled in under the guise as organic endorsements.

It's even worse because LLM providers don't even need to be doing anything malicious for the conclusions to be garbage.

The vast majority of information that the LLM "reads" about any given product is going to come from listicles and other poorly researched "reviews", ad placements, astroturfed comments, and marketing material. They launder all of this together, "summarize" it and present it as rigorous market research. Garbage in, garbage out.

Good luck dealing with the Pink Elaphant problem. Telling a model to not do something in the prompt is one of the best ways to get the model to do the thing.

When billions of revenue are on the line, the teams that OpenAI is currently hiring will spend years to figure out something more clever than my 30 second hack. The example above was a surprisingly effective proof of concept (seriously, try it out), it won't showcase the end state of the LLM advertising industry.

Sure but the assumption here is that the game stays the same. That the only worthwhile intelligence is one that optimizes for revenue capture inside an ad economy.

But there’s a fork in the road. Either we keep pouring billions into nudging glorified autocomplete engines into better salespeople, or we start building agents that actually understand what they’re doing and why. Agents that learn, reflect, and refine; not just persuade.

The first path leads to a smarter shopping mall. The second leads out.