Remix.run Logo
JCM9 3 days ago

This does all smell a bit like when WeWork was buying up seemingly every available office for rent. When it came time to actually pay for said offices… oops.

mike_d 3 days ago | parent [-]

> When it came time to actually pay for said offices… oops

I was at WeWork around the time of its downfall. I have a lot of opinions about how that place was ran, but I can assure you pre-pandemic they were buying up every office space because they were filling them with tenants. Not paying for offices was a result of tenants not paying due to the pandemic.

JCM9 3 days ago | parent | next [-]

And these GPUs aren’t sitting idle either. Nobody is questioning that they “need” the compute, it’s the lack of a viable business model to pay for all this long term that has folks worried.

That’s the same as what happened when WeWork was buying up office space pre-pandemic and then using handwavy nonsense like “Community Adjusted EBITDA” as part of the smoke and mirrors to pretend like there was an actual business there.

The pandemic expedited the pain, but the business model was broken and folks called BS long before Covid hit.

treis 3 days ago | parent | next [-]

>it’s the lack of a viable business model to pay for all this long term that has folks worried.

They're going to sell ads at the moment people are looking to buy stuff. It's the single most viable business model we've ever seen.

jcranmer 3 days ago | parent [-]

I know that "ads" is the popular assumption for how you're supposed to make money, but most sites don't really make all that much money with ads. Facebook and Google aren't rich from selling ads, they're rich from all of the ad infrastructure and quite frankly their ability to follow you around on the internet.

Besides, how are ads on ChatGPT supposed to work? If some student is asking it to write their paper for them, is ChatGPT going to stop in the middle of it and go "Hey, you know what sounds good right now? A nice bowl of soup..." Although admittedly that would make for some hilarious proof of people using AI for things they shouldn't...

vlovich123 3 days ago | parent | next [-]

People regularly ask ChatGPT for product recommendations of all kinds, explicit and implicit.

ChatGPT will also probably be selling ad infrastructure to inject ads just like Google injects ads into search. They probably will pay out little to websites that include the “ChatGPT” widget to integrate ChatGPT with their site that also has ads.

Right now the barriers are technical for injecting ads into AI responses.

XorNot 3 days ago | parent [-]

That directly tanks the entire value of AI though.

As an advanced research engine, knowing it will reliably only recommend you sponsored products means it's worthless - and worse it will be primed to advocate for sponsored products.

Then the whole thing becomes a scam engine, because check out what Facebook ads look like today.

vlovich123 3 days ago | parent [-]

That’s like claiming that “as an advanced search engine, knowing that Google promoted sponsored products and reranks its algorithms accordingly sometimes means it’s worthless”

Regardless of if that’s true, it’s clearly still a huge business opportunity. And you point out Facebook ads are a scam yet they bring in $164B/year and growing. Regardless of value judgement, there’s clearly a lot of money to chase.

treis 3 days ago | parent | prev | next [-]

OpenAI is positioned to be something like Adwords is/was. Think free LLM for your app/website/store or as the backend for products like character.ai. That will let them vacuum up a ton of user data.

Plus like Google search they have a ton of organic traffic. Chatgpt has replaced Google search as my starting point to investigate anything. Lots of that is related to things where I will eventually spend money

jonfw 3 days ago | parent | prev | next [-]

> ChatGPT going to stop in the middle of it and go "Hey, you know what sounds good right now? A nice bowl of soup..."

Google/facebook do that today, because the content they're showing is created pre-ad, and the ads have to be injected after the fact.

With AI- the content is being generated in the same place that the ads are being injected, which allows us to be much more subtle about it.

How much do you think a car company would pay for to put special training weight on their marketing materials? I would guess big money

candiddevmike 3 days ago | parent [-]

I don't think you'd want to do this at the training weight level. That could lead to wildly inappropriate references to your product, potentially to the determinant of your brand.

"While we're on the topic of self-harm, did you know the ABC Co Truck has the highest safety rating?"

WheatMillington 3 days ago | parent | prev | next [-]

Google and Facebook are not rich from selling ads.... that sure is an opinion that you're welcome to have, I suppose.

mossTechnician 3 days ago | parent | prev | next [-]

Those ads would also get mixed into content the advertisers would probably not appreciate.

Zambyte 3 days ago | parent | prev | next [-]

> Facebook and Google aren't rich from selling ads, they're rich from all of the ad infrastructure and quite frankly their ability to follow you around on the internet.

https://openai.com/index/introducing-chatgpt-atlas/

> Besides, how are ads on ChatGPT supposed to work?

"How do I do XYZ?" "Product ABC can do XYZ for you."

burningChrome 3 days ago | parent | next [-]

Which then begs the question if I was the user: "Am I being served an ad or is ChatGPT taking in a bunch of product information and giving me an objective answer to what I'm looking for?"

This would create a ton of hesitation to use this for product recommendations if I knew ChatGPT wasn't using its extensive input for products and reviews and coming back with an objective answer for me.

I guess at this point would we even know the difference? Is it possible this is already happening?

treis 3 days ago | parent [-]

It's what Google does today and they're the 4th most valuable company in the world.

XorNot 3 days ago | parent | prev [-]

Would you use an LLM knowing it will never answer honestly and any recommendation is likely a sponsored one?

kemotep 3 days ago | parent | next [-]

Hell is it going to start injecting ads into coding output? Ask Codex to generate you a fix for your web app and it spits out a number for a web hosting service? Give it a Jira ticket and it gives you an ad for a different SaaS ticketing system?

Is it going to inject ads for indeed while a recruiter is using ChatGPT to summarize a stack of resumes?

If it only ever injects ads for specific requests how profitable would that even be? I understand clients would want their product to be recommended but if I only get the ad answer when prompting a certain way, can I the user avoid ads by asking questions a specific way?

Zambyte 2 days ago | parent | prev [-]

I don't matter. Will a substantial portion of the population? Unfortunately it's not something the average person makes an active decision about, understanding the situation.

dktp 3 days ago | parent | prev [-]

I am also in the camp believing they will sell ads the second they find a viable way (churn worth it, base infrastructure for it built, enough people trusting ai with product recommendations...)

I think the queries will fall into profitable (product recommendations) and non profitable (writing an essay or code) just the way they do for Google. Probably former will have a generous free tier and latter will be largely paywalled. I don't know how they'll do that, but I imagine they'll find some way

It's a mass consumer (software) product and they need new revenue venues and ads have a history of working well. Even Spotify, Netflix, Amazon Prime, ... Companies that historically don't have the ad infrastructure of Google or Facebook have increasingly profitable ad tiers

mike_d 3 days ago | parent | prev [-]

I was simply addressing the implication that WeWork was just buying up office space for fun and not paying for it.

asah 3 days ago | parent | prev | next [-]

WeWork was taking on long term liability commitments and paying for them with short term revenue commitments. One bad thing and poof. Everybody in the commercial real estate market saw this coming.

OpenAI maybe in the same situation, committed to spending $1.4T while enjoying a good revenue year this year but then One Bad Thing and poof.

rchaud 3 days ago | parent | prev [-]

WW was able to fill those offices by charging well below market prices, because they were VC-funded so growth was more important than profitability. OpenAI is doing the same. 800m users, the vast majority of whom are free users who won't convert to paid.

winstonp 3 days ago | parent [-]

won't? i doubt that. can't tell you the last time i paid $20/mo for any non-business sub, but i've paid for some combo of openai+claude for last year