- Novel way of introducing and learning more about sponsored products
- Strong branding for non-techie people (most normal people don't know what Claude or Gemini are)
- An app that is getting more and more addictive/indispensable
I think OpenAI is going to kill it in ads eventually. This is why Meta and Google went all in on AI. Their lucrative digital ad business is in an existential threat.
I think people who kept saying there is no moat in AI is about to be shocked at how strong of a moat there actually is for ChatGPT.
All free LLM chat apps will need to support ads or they will eventually die due to worse unit economics or run out of funding.
PS. Sam just said OpenAI's revenue will finish at $20b this year. 6x growth from 2024. Zero revenue from non-sub users. What do you guys think their revenue will end up in 2026?
> most normal people don't know what Claude or Gemini are
In think the point is that they don’t need to know what Gemini is, they just need to know Google, which they most definitely do.
IMO ads rollout won’t be as simple as you’re describing it. A lot of people have switched from Google search to AI specifically because it isn’t filled with SEO, ad filled nonsense. So they’ll need to tread very, very carefully to introduce it without alienating customers. Not to mention mollifying advertisers who are nervous what their product will be shown alongside and OpenAI will probably struggle to offer iron clad guarantees about it. And people generally speaking don’t like ads. If competitors like Google are able to hold out longer with no ads (they certainly aren’t wanting for ad display surfaces) they might be able to pull users away from OpenAI.
IMO pivoting to ads is a sign of core weakness for OpenAI. Anyone trying to set up their own ad network in 2025 has to reckon with Google and Meta, the two absolute behemoths of online ads. And both also happen to be major competitors of OpenAI. If they need ads that’s a problem.
Poe's Law notwithstanding, I find it hard to believe that anyone would think I was making a good faith business acumen observation. If Optimus walks you to the kitchen to get a coke, what's Tesla's business model? Charge by the nanosecond for compute time?
Purchase/lease access to the hardware, subscription for the necessary online connectivity, and microtransactions for each actual use of it (ostensibly because of cloud compute, and that also means surveillance data is captured and monetized).
The business model for Tesla and xAI is actually very simple and superior to OpenAI and Google's. No, this is not satire:
The business model is that his companies are meme stocks, and controlling social media means controlling meme stocks. The business model is also that his companies require corporate socialism, and controlling social media means influencing government policy.
He can talk about AI driving cars, but that's yesterday's news. Today, his business model for AI is to put his finger on the scale and influence society to help him become richer. AI is threatening to replace search, but in a way it's also threatening part of what social media provides, namely the ability to guide discourse at scale.
What's easier: Getting his personal board to give him a trillion dollars, and shoring up public support for that with bias in his AI products and on X? Or building a trillion-dollar business?
Elon Musk's business model for AI is actually quite easy to understand.
It wont be that obvious. It will explain to you the dangers of doing your own cooking, the number killed by food poisioning each year, then suggest something from doordash instead. Or it will suggest you eat something faster, like pop tarts, so you can spend less time cooking and more time interacting with your AI buddy.
> So they’ll need to tread very, very carefully to introduce it without alienating customers.
every single platform since the 1990's has introduced ads. My kids find it totally normal to have them. Believe me, if you train (!) people to accept ads, they will soon think it's normal.
And besides, if ChatGPT goes with ads, Google will follow directly. So the users won't have the choice anymore.
But ok, if I have to pay for a service without ads, then let it be. Paying for a service is normal too.
> So they’ll need to tread very, very carefully to introduce it without alienating customers.
I'm certain the ads will be introduced in an easily identifiable and ignorable way. People will acclimate, user behavior will be analyzed, and over time the dial will ever so slowly be turned up to optimize for draining as much attention and money from the consumer as possible.
You'll just need to run a small local model to filter out the ads. And they just become another one of those silly arms races between the ad makers and the ad blockers and we all burn more electricity.
> IMO pivoting to ads is a sign of core weakness for OpenAI.
Yeah, I've had the same thought for a while now. You don't sell investors on an endeavor for 10s of billions of dollars with the endgame being "sell ads". If that was the endgame then there are a lot less resource and capital intensive ways to get to it.
Given all of the discourse of "you need this new tech in your life to continue to participate in society", I would not have expected them to need to stand on the roadside trying to get people to buy low cost fireworks. It smacks of going through the sofa for loose change so you can make rent.
And if they had something impressive coming down the pipeline I would think they could get someone to spot them a few billions yet, unless the billionaire/megacorp economy is really that tapped out.
I am not exactly a great example ( exposure to work model, ollama, local models play ) and I actually liked gemini upon try in google search ( which is amusingly now banned at work ), but the nice quickly fell into not nice, when it started giving me weird pushback on operation paperclip book ( I am assuming chapter discussing tabun triggered something. This is my only real problem with gemini. By comparison, I am not running into guardrails with gpt nearly as often.
Chatgpt is a proprietary eponym[1], like kleenex, or Google for search. That's a relatively strong attractor based on their first mover status. I nevertheless use tissues, and search engines like brave search, sometimes duckduckgo, and claude or openrouter for my LLM models.
I think there are too many good alternatives for Chatgpt to turn the screws too hard on their users, but we'll see where it settles out. As usual, the most vulnerable will be squeezed the hardest (the ignorant and tech feeble). Hopefully competition and some oversight will keep the wolves at bay.
The finance people were chatting about the OpenAI's ad play a while back, glad to see it finally dawning on this crowd.
1. Not all jurisdictions have granted OpenAI the Chatgpt trademark.
Weirdly, I think Perplexity is getting a lot of mainstream name recognition because of podcasts. All the big slop pods like Rogan, Theo Von, etc are sponsored by Perplexity and the hosts constantly name check it by asking to “look stuff up on Perplexity”. Honestly pretty smart marketing all things considered.
Claude has been aggressively advertising on Facebook, Instagram, and Reddit, and the ads have been much more general use than just the code benefits. They’re definitely no ChatGPT, but they’re not an unknown player.
You're only seeing those ads because the ad algorithm knows you. My family aren't getting Claude ads. They wouldn't know the first thing about it even if it were explained to them.
Yeah my father who codes occasionally asked me what the best AI for coding was and he had never even heard of claude so I would be very surprised if your average person knows it.
they just need to know Google, which they most definitely do
The way that Google is rolling out AI is confusing, and I imagine a lot of people who can access Gemini don't actually know they can or how to use it. Among those that do know, many won't know what it's capable of and will believe that they need to pay for a service like ChatGPT in order to get what they want.
And yet most of the people I know, including many technical ones, default to ChatGPT before Google's AI Studio. Google has general brand awareness, but ChatGPT has become the Bandaid or Kleenex of AI
I agree but how many consumers actively purchase Bandaid or Kleenex over cheaper store brands? Becoming a generic term doesn’t always translate to great business. “I’ll put it into chat” could easily end up meaning “enter into Google’s AI prompt” for many people.
I think investors would certainly love this. So why hasn’t it already happened?
My guess: they would lose a ton of cultural cachet.
Turning OpenAI into an ads business is basically admitting that AGI isn’t coming down the pipeline anytime soon. Yes, I know people will make some cost-based argument that ads + agi is perfectly logical.
But that’s not how people will perceive things, and OpenAI knows this. And I think the masses have a point: if we are really a few years away from AGI replacing the entire labor force, then there’s surely higher margin businesses they can engage in compared to ads. Especially since they are allegedly a non-profit.
After Google and Facebook, nobody is buying the “just a few ads to fund operating costs” argument either.
Yup, it’s essentially an admission of failure. I think the people who were expecting AI to improve exponentially are disappointed by its current state, where it’s basically just a useful tool to assist workers in some highly specific fields.
Highly specific fields? They are trying to get you to reach for ai when an emailed “ok, thanks” would do. They want you to lose your ability to write and formulate thoughts without the tool. Then it is really over. That is the golden goose. Not a couple data scientists.
Yes and there are layers. Remember when google ads had yellow backgrounds? I'm sure OpenAI will find a way to do ads "ethically"... for a while, until people get comfortable, and that's when they will start to make ChatGPT increasingly manipulative.
Can anyone explain to me what ChatGPT does that traps people? I get the value as tools, I like using copilot, but ChatGPT doesn't offer me value that any other LLM can't. Given that everyone is quickly rolling "AI" into their own stuff, I don't see what's ChatGPT's killer app. If anything, I think Gemini is better positioned to capture the general user market.
The branding is so strong and it works well enough (I’d say, according to the perception of most people) that it’s just the first “obvious” choice.
Akin to nobody getting fired for choosing AWS, nobody would think poorly of you using ChatGPT.
I don’t think Claude has that same presence yet.
Google has a reputation for being a risk to develop with, and I think they flopped on marketing for general users. It’s hard to compete with “ChatGPT” where there’s a perceived call to action right in the name; You don’t really know what Gemini is for until it’s explained.
> When was the last time you went to an actual physical library
My eyesight is making paper books harder and harder to read, so I don't go to libraries and bookstores as much as I used to. But I think libraries are still relatively popular with families, because they're sites of various community activities as well as safe, quiet places to let kids roam and entertain themselves while the parents are nearby.
When I was a kid, my parents went to the library much more often than they do now, because they were taking me and my sister there. And then we would all get books before we came home.
Not saying you're entirely wrong, but there's a significant part of this that is "changing rhythms of life as we age", not just "changing times".
It used to be, people went to the library to look things up, and as a primary source for finding information they needed. Not just as a community center.
> Gemini is a competitor, yes. But most people still go to Google at this point, even if there are a ton of competitors.
Yeah, that's my point. If Google is good enough I don't think people are going to want to do those extra steps, just as in your google maps example. There might be better services out there, but google maps are just too convenient.
Only a short matter of time before agentic tools start serving ads too - paying user or not. You want to refactor your codebase ? No issue - taking 30 seconds - please view this ad meanwhile.
In my head, it'll be like the high pressure timeshare sales pitches or the dreaded car sales transactions, where they pull out all the tricks to convince you to buy something you don't actually want or need, regardless of whether you can afford it.
“That’s a great point about your finances. But did you know this company offers credit to someone in your position for only this low interest? I can apply on your behalf if you just sign this statement”
It’s worse than that. It will be more akin to ad placement in movies, except in this case it will slip this proprietary library into the suggested solution instead of that one. Or embed ads in the solution code.
eh, maybe for the super vibey tools. I can't imagine anyone who wants to maintain the trust of devs would do this, they would so quickly pivot to something else, its not like the general public, where AI == chatgpt in their mind.
Why stop there? You want to refactor your codebase? Sure, but you need to adopt this dependency and this cloud service.
Mixing Ads or sponsorships to influence LLMs is a really, really bad idea. Especially when they're competing with Search ... which means that for some, "AIs" are the only window into the world when looking something up.
> Sure, but you need to adopt this dependency and this cloud service.
THIS.
Asked to make an app using AWS? “I can do that, but have you considered the lower lifetime costs of using Azure? I can generate a configuration for AWS, Azure, or produce a price comparison table. Let me know which you prefer.”
Absolutely no business with any real money is going to be swayed to use one cloud provider over the other based on one vibe coded app.
There is so much other stuff that goes into why business make decisions about any large contract. I’m not in cloud sales. But I venture just close enough to the sun not to get burned
> Absolutely no business with any real money is going to be swayed
And what about everyone starting something? Or prototyping? And what if you don't have a choice: pay more or follow our sponsored guidelines?
This is a dangerous road without proper defences both in terms of legislation and policy (and I mean world-wide, world corps = world laws, not having to go to court in every country lol).
Also, end users need to be educated about all this because what is to stop John or Jane from uploading their receipts to GPT to make their taxes and ... oops "did you know you can switch insurance to XYZ" or ... AI browser proactively hiding content competing with their partners ... you looking for a healthcare package? The only one available is from our sponsor. Take it or leave it.
The tax situation already happens now with Intuit owning both Turbo Tax and CreditKarma to get you to sign up for credit cards.
If I were prototyping something and found I could do it cheaper somewhere else, I’m not sure I would be upset. I hate ads just for the bad user experience.
As long as it is clearly an ad and they say they have affiliations. It’s no different than what Google and Amazon does now.
But ironically enough, I was almost about to pay for Overcast years ago even though the author openly admitted that you didn’t get much of anything for it except for supporting him back then.
He then added a non slimy self hosted system to buy ads for other podcasts based on the category of podcast you were listening to at that very second (no tracking). I thought that was a great service.
I think I would actually lean into a tight integration between ChatGPT and something like booking.com[1], AirBNB, GetYourGuide, etc when looking for travel ideas.
[1] Well I personally wouldn’t because I am not as cost conscience as the average traveler and I value the loyalty programs and status of certain hotel chains and Delta airlines. But most travelers don’t and shouldn’t care.
But if they let me put in my loyalty numbers and book directly with Hyatt, Hilton and Delta, hell I might pay more for ChatGPT.
How many medium size migrations have you done? Its never that easy even if you are just hosting a bunch of VMs. Let alone if you are using any cloud specific services.
I have been involved in a few on the periphery working in cloud consulting (first at AWS itself and now an outside company). I actively avoid the “lift and shifts”. I come in for the “modernize” portion.
I have been either part of or opted out of well over a dozen - I have a policy of never leading “lift and shifts” (and never staff augmentation).
Once you actually sit down and come up with a project plan with your PMO, a cloud migration is hardly ever worth the effort unless the destination cloud provider is backing up a shit ton of money for not only operational credits but also for internal AWS Professional Services (where I worked when I was inside AWS) or an outside partner to help (current employer).
Hell even certain departments at Amazon would never go through the effort of migrating to AWS from the legacy CDO infrastructure.
The risk of regressions, the refactoring, the retraining, the politics, etc are hardly ever worth it.
I highly doubt this one. These agents are pretty interchangeable, any one of them could decide to not show ads and steal huge market share. Programming is one of the few areas they can actually make gross margin, so ads would be a terrible business decision given the above. The ad revenue from it would be insignificant against the API calls / subscriptions.
"I think people who kept saying there is no moat in AI is about to be shocked at how strong of a moat there actually is for ChatGPT."
I'm not sure that really is the case. Most non-techies I know use ChatGPT far less than they use Google search, let alone various social media apps they're addicted to.
Perhaps it is a threat to Google search, but I can't see how it's going to be threat to ad revenue from Meta, Youtube etc - the services that are actually addictive due to the content they serve. At least for me there's absolutely nothing addictive about ChatGPT. It's just a tool that helps me solve certain types of problems, not something I enjoy to use.
> - Extremely personal data on users
- Novel way of introducing and learning more about sponsored products
Doesn’t anyone think this is really, really bad idea? We managed to radicalise people into the rise and fall of entire countries through analog ads, can you imagine how devastating it would be to infuse every digital product with all that?
Also appeal to investors. Nobody would give tons of money to upstart which goal is to generate text porn, generated TikTok slop and make some needy teens suicide just to compete with Google Ads.
Selling big AGI dream that will literally make winner take it all is much more desirable.
You know, I thought stories of law enforcement and the military targeting people using commercially collected data, effectively skirting the sanity boundaries we applied to surveillance, would raise a little bit of awareness in the US. It didn’t. Then when the political scene got really into deliberately targeting political opposition, I thought that might raise more eyebrows about all of this data being out there, but it didn’t. Same with ICE and border patrol. I think the risk and mechanisms will remain too abstract for people to grasp until they’re one of the unlucky people staring down the barrel of a gun because they, or someone they were associated with, had the wrong opinions.
This has echos of “First They Came” [0]. The current status quo begs a question that must have been asked in the time it was written: at what point does it have become morally acceptable for citizens to rise up and overthrow a violent government?
The current dynamic IS based on the new gestapo thinking they're "rising up and overthrowing a violent government", due to a propaganda bubble thirty years in the making. Where do you think "ICE" is getting all of these new recruits from? The red state militias that have been seething about the slow creep of bureaucratic authoritarianism, now deputized and told they get to use their weapons to attack the other tribe. Which is also why said militias are silent now that it's actually time to defend our country from fascists - they are the fascists. Propaganda is a hell of a drug.
You seem to be stating this like I said something that might imply otherwise, but I can't figure it out even seeing you've got the GGGP comment. This thread kind of went off on a tangent that isn't directly addressing your original point.
But to hopefully answer your question - yes I'm in favor of wholesale importing the GDPR as-written into US law and letting the courts sort it out (sidestepping the corruption^Wlobbying process wherein corpos would make "small" edits that effectively gimp it with loopholes). I'm also in favor of antitrust enforcement against companies that anticompetitively bundle software with hardware and/or services - ie people should be able to choose software which doesn't have ads, rather than being coerced by the pressure of network effects. And if neither if those were enough to stamp out the consumer surveillance industry (aka "Big Tech") as we know it, then I'd support directly banning personalized advertising.
(I would support directly curtailing government from abusing commercial surveillance databases as well, but I don't see a straightforward meta-way to prevent that besides drastically shrinking the commercial databases to begin with)
It’s not that they don’t care - the current administration is targeting people that they specifically don’t like.
And Trump has a cult of personality where many Republican politicians are literally afraid for their lives if they stand against him because they get death threats.
Romney said other Republican politicians won’t stand against Trump because they can’t afford security like he can. Majorie Green Taylor said her family has started getting death threats and the Indiana legislators who were first opposed to redistricting are now holding a vote because they also got death threats
I just got back from Thanksgiving holiday with my family. Grade schools kids all the way up to great grandparents up to 81 years old. Engineers, active military, a nurse, high schoolers, two in college. Both coasts represented and Texas. Republicans, Democrats, and in-between. The one and only thing every single person had in common was an utter hatred of AI. And it wasn’t for a lack of understanding of how it will be used.
There are so many more reasons to hate AI than just “it is taking my job”. But even if we’re just sticking to that, some people don’t like that it will replace their co workers, neighbors or family members job.
1. The absolute explosion of AI usage (revealed preferences)
2. The polling on AI, which is mixed and reveals lots of pessimism and fear among a slight majority of Americans, but hardly universal “utter hatred”.
My guess is some combo of: your family is not representative, the hatred was not as universal as it appeared (bandwagon effect), or your own hatred of AI caused you to focus on the like-minded opinions shared and ignore any contrary evidence.
Netflix never introduced ads in the ad free service. They introduced a new lower tier price with ads that if you were an existing customer, you were none the wiser.
> They introduced a new lower tier price with ads that if you were an existing customer, you were none the wiser.
You're right that I didn't experience them myself, but my data here are (1) Netflix evidently getting a lot of takers and making a lot of money from people using this new with ads tier, and (2) the lack of any sustained negative outcry against Netflix after the first news cycle or two.
So I'm intending to rely on that rather than my own experience. OpenAI has any number of permutations of ways to include ads, including a Netflix style cheaper paid tier, so I don't necessarily think a distinction holds on that basis, though you may be right in the end: it's more intuitive to think OpenAI would put them in the free version. Though it's possible the Netflix example is teachable in this case regardless.
> about Netflix doing it, but they did and the world just moved on
I think the main challenge here is that Netflix works around one of many ways to access entertainment. So if one service starts to show recommendations in that limited context of user data they collect - it's still has negative potentials but it's easier to regulate and there are alternatives.
In the case of LLMs, we have service that are aiming to replace both the browser and the search engine. This means ending up in a situation where your entire access to knowledge and the world takes place via "AI". And the result is: ad-infused, tweaked to align with investor priorities, censored by the current politics of wherever the company is based service machinery that's constantly extracting personal information so it can learn better ways to refocus its priorities. I've read and seen a lot of sci-fi and dystopian history novels (actually read, not LLM-summarized for me) to know this is a very end-game kind of situation.
It's one of the major issues of our era. Either society will be utterly captured, gradually and quietly, or there will be a reconning and ads will become tightly regulated along the lines of tobacco, sectioned off from polite society.
Spinning up an all-new ad network is pretty tough. I would think OpenAI would need to beat Meta/Google on basics like CPM in order for the network effects to make it desirable for ad vendors over Meta/Google. Ad budgets are fixed and zero-sum and vendors (in my head, I don't know) would prefer to spend their money on the best network giving the best results. I don't know if ads in LLM chats can get there.
ChatGPT shows a sponsored entry in chat history list with a colorful border around it to get users to click. This product is something that ChatGPT knows the user desperately needs from previous chats. The user can chat directly with the product and learn more about it. The advertiser specifically sent OpenAI information (like a RAG) about their products buyers might have questions for.
When the user is ready, they can open a link to the product's website or just buy directly in ChatGPT.
But the missing part is "we know they need this!" but they don't have the ad network to have the pixels on the vendor sites to track the conversion (or not for remarketing!). They only have half (at most) of the picture. This is why they tried to create a browser (remember that? Nope me neither) to try and get the full picture.
Advertisers are accustomed to pay for conversions now. If you can't track it, you cant prove it happened.
Open ai will need to spin up the entire infrastructure (Inc sales teams, support teams, servers etc) to run the ad network. Not impossible but it is a big lift and they're already burning money.
Their best bet is probably to just sign up for selling their ad space with Google, like all the other apps and websites do
I think the amount of money they’re burning on their operations would make that organizational lift a drop in the bucket. A few dozen annual 6 figure salaries? A few hundred? A bunch of normal CPU-based AWS services? They must spend 10 million per day on their current operating expenses.
Google and Meta are many thousands of sales people, managers, engineers, SREs, HR, masseuses etc. If you want to scale to Meta or Goog scale when doing ads you won't be able to do it with a few dozen people. Just sales will be hundreds or thousands spread across all the major territories
Except OpenAI’s value is overinflated by an economic bubble. They don’t have the manpower or resources to effectively implement an ad network on the scale they’d need to become profitable. Why are you insisting we keep the conversation positive?
They can also suggest chat topics. Like how Reddit ads are meant to look like threads. "Frustrated with food baked on dishes? Let's chat about it! (sponsored)"
I thought that most advertisers go through middlemen and do not do business with the ad networks directly? So you only have to make it attractive for the middlemen (of which there are fewer), and that shouldn't be a problem for anything AI-related.
Furthermore, anyone offering some sort of assisted browsing service is automatically in the ad business, regardless what they do with affiliate links in generated page summaries.
Oh yeah and on top of that these companies like WPP hate the fact that Google and Facebook refuse to share more information with them. They can't wait to jump ship.
They have all the resources anyone could possibly need to do this, including an enormous list of companies who would kill to get their products into ChatGPT. It’s “just” an execution challenge.
Spotify, Netflix, Amazon Prime, Reddit, Twitter etc all have increasingly profitable ads
I'm sure llm providers will also figure it out in due time. Consumer products are generally a good fit for ads, even if it takes time to reach full potential
Every single one of those companies have ridiculously low marginal cost per request compared to ChatGPT and much lower fixed costs and continued development costs.
Why would they need to beat Meta/Google, and at what game? They just won’t let any other add network work in their app. Voila! You just beat Meta/Google, and they didn’t even compete with them. I guess they could provide some sort of SDK for websites to embed that tracks users, or they could come up with a browser extension that tracks users too. They already have an app that people are freely giving them so much info. Where else could the compete as you suggest? Being a generic ad platform to serve ads not through their app?
If OpenAI manages to get the agentic buying going, that could be big. They could tie the ad bidding to the user actually making the purchase, instead of just paying for clicks.
Why would OpenAI want to use and pay for any of Microsoft's products unless mandated by a contract?
OpenAI has the talent to roll out and run their own ad product that is better and more efficient. Why pay Microsoft for a core part of their (future) business?
P.S. In case you haven't noticed, OpenAI demos are done on Macbooks. Microsoft could not even get them to use Windows.
Microsoft runs many of their demos on MacBooks. You missed the memo that Windows OS is no longer their bread and butter. Go check their GH OSS projects (e.g. .NET) and all of them have have shell scripts alongside PowerShell.
This reminds of ads on Amazon Echo and other intelligent devices. I think there was similar hype - not in terms of scale - but on personal data. Many advertisers and their moms were writing skills to tap this market.
It'll be interesting to see how they serve up ads and how it ends up working. Before the initial state is that people will find ways to serve up malware in form of ads and someone might end up writing ublock type stuff to block these ads.
I’m actually one of the people that continue to say even with this list they have no moat, because Google, Facebook, Microsoft, etc. can just embed a chatbot in their existing products or social network and make ChatGPT irrelevant overnight. Non tech users will chat through their browser, OS, Apps, website, that’ll be served by any model provider. The only moat of OpenAI is investor money to burn so that they can offer it for free.
Also 20 billions of revenues, not profits, is orders of magnitude too low compared to their expenses. Their only path to survival is a massively downgraded free tier ridden with ads. Nobody will use an app like this when they can have a better more integrated experience directly in their other apps.
> Also 20 billions of revenues, not profits, is orders of magnitude too low compared to their expenses.
Nah, it's just one order of magnitude...
Also, they expect revenue to grow exponentially so it's 20 billions annualized by the end of the year. Last time I saw somebody talk about it, it was about half of it, and trending down.
Anyway, if they manage to take ~20% of the ads revenue from Google, they will be able to cover ongoing depreciation! That's the amount of money they need.
The problem is that inference is a whole different ballgame in terms of costs compared to a traditional SaaS model where each extra customer adds near zero in cost.
They may make it work but OpenAI is more akin to a traditional high revenue low profit business like for example a grocery store.
Thats why we are seeing the explosion of extra tools to try lock in business for higher value use cases and not fight on the margin.
It’s an entirely different skillset to create technology $x than it is to create a successful ad network. Yahoo is the canonical example. It has been one of the most trafficked websites in the world through most of its history and still wasn’t able to successfully sell ads after the dot com bust.
ChatGPT’s revenue means nothing if reports are to be believed that it loses money on each paying customer on just inference. It’s definitely not enough to support its training costs.
Also, I think I remember estimates that it costs 10x as much to serve a ChatGPT result than it does for Google to serve a search result. Not to mention that Google uses its own hardware including TPUs.
ChatGPT’s revenue means nothing if reports are to be believed that it loses money on each paying customer on just inference. It’s definitely not enough to support its training costs.
> I think people who kept saying there is no moat in AI is about to be shocked at how strong of a moat there actually is for ChatGPT
Given one can (at least for the moment) export one's entire chat history from ChatGPT, what exactly would stop a ChatGPT user from switching to an alternative if the alternative is either better, or better value?
The answer is friction. What % of this billion of users will bother to export their chat history (which is already a lot) and import another another llm. That number is too small to matter.
How many of those care about their own particular history in the first place and what % of those at least actively manage it outside of standard chat interface or even hop providers? I think that % would surprise you.
All chat apps look exactly the same and have exactly the same features. The friction is basically non-existent compared to email services, social media, web browsers, &c.
I think it matters to more than you might think. A significant portion of the non-technical ChatGPT userbase get really attached to the model flavor.
The GPT-4o controversy is a good example. People got attached to 4o's emotional and enthusiastic response style. When GPT-5--which was much more terse and practical--rolled out, people got really upset because they were treating ChatGPT as a confident and friend, and were upset when it's personality changed.
In my experience, Gemini and Claude are much more helpful and terse than ChatGPT with less conversational padding. I can imagine that the people who value that conversational padding would have a similar reaction to Gemini or Claude as they did to GPT-5.
Yet, somehow I've been paying $20/month to ChatGPT for years now and I don't use Claude or Gemini even when they're free or have slightly better models.
Not having used anything except for Firefox, I don't have any experience with migrating to different browsers. However, my understanding is that Chrome shows a little pop-up that lets you import from previous browsers rather than relying on the user to do a data export. Correct me if I'm wrong about this.
I suspect that Claude couldn't make an "import from ChatGPT" button because OpenAI would make it difficult, so they'd have to rely on user initiative and technical capability (exporting to JSON and importing from JSON is enough technical friction that the average user won't bother).
Try telling your PM that you want to ignore Safari when you create your website with 60%+ of mobile users in the US using iPhones and globally your most affluent users are on iPhones. Even if they download Chrome for iOS, they are still using WebKit.
You are vastly overestimating people's willingness to deal with bullshit, when the product does not have a real lock in.
It would be incredibly easy to have a company offering their ChatGPT over WhatsApp or iMessage, and get people to start using it instead of an ad-ridden GPT app.
People are being weird about this. ChatGPT has no moat because switching costs are zero. There's no investment into a particular AI service.
ChatGPT has mindshare but that's not the same as it being a moat. The fact that people will continue to use ChatGPT after some gentle frog boiling is true of any service. Adding ads is going to be a measure of how real people tolerate ads more than anything about ChatGPT. Normal people really don't care that much and it bothers me—and probably most of HN.
And the ads can be blended seamlessly into generated content.
"You can do this in Postgres, but the throughput will be limited. Consider using hosted clickhouse instead. Would you like me to migrate your project?"
Hard disagree on the moat. I do tech diligence on "AI startups" regularly and so far have yet to hear of one that has had a hard time ensuring they can use competitors just as easily. Everyone is very aware of that issue, if blissfully ignorant of others.
Also ads in LLM can be perfectly merged with the content, it'd be impossible to know if LLM tells you something because that's the most likely useful answer or the most profitable one for its owners. Can't be just ad-blocked either, it might be the ultimate channel for ads.
> how strong of a moat there actually is for ChatGPT.
None of the above requires OpenAI to be around though. Google, Apple and Microsoft each have much stronger brands, and more importantly they each own large platforms with captive audiences where they can inject their AI before anyone else's and have deeper pockets to subsidize its use if need be. Everywhere OpenAI opens up shop (except for Web) they're in someone else's backyard.
And include an ad section within the text. Alternatively, if it tells you something because that company is a sponsor, it could just include an appropriate disclaimer.
I think ChatGPT’s moat is mostly “it’s the first AI thing I used/heard about”. It’s not clear to me that’s enough to maintain their market share if OpenAI is the only one mixing in ads. It does seem to work elsewhere, though; consumers have brand loyalty to a fault, and often for the brand they started with.
The question is how many users have developed intimate personal relationships and have named their ChatGPT, and how many of them would bounce to a different provider if some line is crossed (of which advertising could be one)?
Anecdata isn't data, but I know several individuals who have and thus are even more unlikely to churn than mere brand loyalty on the level of eg Coca-Cola.
Google, Microsoft, Meta and Amazon, among others, would have zero issues in ensuring that OpenAI does not grab a market they own; it shouldn't be that hard to bring OpenAI into a position where they cannot recoup their investments, hence going bankrupt.
The big players then would also have the benefit of having those very bright minds being on the market for them to grab. And it's not like OpenAI owns much relevant hardware.
I do not understand why the conversation is always about showing ads in chatgpt. Can they not track users there without ads and sell ad space on websites like google ads? Why ruin the experience there when they can highly target ads. I am guessing they prefer both.
There is nowhere near enough money in the ad network business. Like, Google's search ad business is an order of magnitude higher than the ad network, and the ad network has been shrinking in absolute terms for years while the first party revenue has been growing at double digits.
Eh, so.. I don't know if I was in some weird A/B testing group, but I saw lazy reference to real estate ( zillow ) in my chat few weeks ago, which was .. I had to think of a way in how 'not close' it was to our conversations. And the issue is clearly not that it can't profile me. It absolutely can. And I sometimes ask for some explicit shopping comparisons. But what do I get, real estate ad..lazy. Lazy and uninfuckingspired.
I think the good news is that open-source models are a genuine counterweight to these closed-source models. The moment ads become egregious, I expect to see and use services for an affordable "private GPT on demand, fine-tuned as you want it"
So instead of a single everything-llm, i will have a few cheaper subscriptions to a coding llm, a life planning llm (recipes, and some travel advice?). Probably it.
They can easily LoRA-finetune each model based on user preferences expressed in the past conversations. That would improve accuracy compared to Google's ad targeting by orders of magnitude.
I think the real question is: what are you doing to make it less painful? Full disclosure, chatgpt has a lot on me, but I am using to this time to prep nice local build. It has gotten really nice and current crop of machines with ai395 got really nice ( I almost wrote a short page over how easy it was compared to only few years back ).
Hopefully this is the point where AI starts to be seen as just a useful tool, as opposed to a sign of imminent AGI. I’ll be glad to hear less rabidly overzealous rhetoric surrounding AI.
How many of those are actual active users though? I created my account when chatgpt 3.5 was launched because it was a novelty but haven’t used it in a long time. I use Claude and Gemini but I’m somehow counted in that 1 billion figure
Sam is a habitual liar so I wouldn’t take anything he says seriously.
LLMs are a commodity, once they put in ads people will increasingly move to the other options. It works for Google because they have a moat, OpenAI does not.
There’s a reason they didn’t do this earlier. It’s going to piss people off and they’ll lose a lot of users.
Trust in LLMs is easily broken, and many users are starting to see the cracks. Once those AI companies start rolling out ads inserted in the answers, the quality will go down even more, and they will burn the last good will of the people.
There is no moat because their only way to make money is to self-destruct.
Talking on a more practical POV, your cost to display the ads needs to be lower than what companies pay you for advertising. And while companies might be willing to pay a small premium for "better" targeting because the LLM supposedly has more personal data about users, the cost to deliver those ads (generating answers via LLMs) is several orders of magnitude higher than for traditional ads served on websites.
So even sticking to a purely technical aspect, ads might simply not be profitable when integrated in LLM answers.
Combine the two aspects, and OpenAI is all but a dead company.
In this niche forum people keep saying “there’s no moat”. But the moat is the brand recognition, if I ask my 70yo mum “have you heard of Gemini/Claude” she’ll reply “the what?”, yet she knows of ChatGPT.
Does Coca Cola have a moat? Some company could raise $1B to create a new cola beverage that beats Coca Cola in all blind tests imaginable yet people will keep buying Coca Cola.
Did people switch search engines or social networks when Google or FB introduced ads?
I wouldn't call ChatGPT "brand recognition". People know the term ChatGPT, but I don't think they associate it with OpenAI or any company in particular, in the same way that people might associate Civic with Honda. Instead they'll associate it like they do the terms Bandaid, Kleenex, etc., as a catch-all term for LLM chat interfaces, regardless of who is providing the service. When OpenAI starts ads, I imagine people will start saying "oh, here's a ChatGPT without ads" and point to Claud or Gemini or whatever.
> But the moat is the brand recognition, if I ask my 70yo mum “have you heard of Gemini/Claude” she’ll reply “the what?”, yet she knows of ChatGPT.
Brand recognition doesn't mean a thing when it comes to a technically-illiterate audience with no control over their digital lives. In the same way that every 90s mom called a video game console a "Nintendo", everyone who gets served an LLM-generated response straight from their OS and/or browser courtesy of Google, Apple, or Microsoft will call that a "ChatGPT", and OpenAI will be powerless to stop the platform holders from intercepting their traffic.
I don’t disagree but want to go on the record predicting this will collapse on itself spectacularly and OpenAI will still “fail” commercially
for the Cola Cola drinkers, the product goes from an infallible AI to with no ulterior motives to another Google that’s purpose is to sell you ads, but more creepily. it’s like if Coca Cola started adding a few milliliters of bleach to their product
This isn’t backed by the constant conspiracy theories about voice assistants listening to everything you say and then farming that off to third party ad providers so that you see ads for things you’ve been discussing.
I’m not certain about that, but it’s all very abstract to people. It is also tied to their phones for most people which they’d never give up anyway.
The more direct connection on something they don’t (yet) value as much as they value their phones might be a bridge too far.
An LLM feels like a person to a lot of people. It might be surprisingly difficult to avoid people feeling betrayed or creeped out by this “person”. No one has ever done this before and it doesn’t seem easy or like a straightforward win.
I think it’s more reasonable to consider Coca Cola as having a significant brand value moat, given that they’re 140 years old and one of the most recognizable brands in the world. That also gets at the other side of their moat: distribution. Coca Cola is available basically everywhere, and a challenger would have to invest massively to simply get in front of as many people on shelves. In that way, other companies (Google, Microsoft, Meta) still have significant legs up on OpenAI. Way too much in play right now to declare any winners.
Don't ask them if they know the model name, ask them if they've used the ai mode in Google search or their phone or Gmail or whatever. "Oh yeah I use that all the time!" is what they usually say to me.
People say ChatGPT has brand recognition but amongst non-students and non-tech in the UK I don't think it is that pervasive at least.
Tried and true Silicon Valley strategy: burn VC money to build a moat, wait until switching costs are high enough, and then enshittify the product to extract rent.
> I think people who kept saying there is no moat in AI is about to be shocked at how strong of a moat there actually is for ChatGPT.
Game on. The systemic risk to the AI build-out happens when memory management techniques similar to gaming and training techniques that make them usable reduce the runtime memory footprints from gigabytes to megabytes, much of which fits in L2. When that happens, the data center will bleed back to the edges. Demand will find its way into private, small, local AI that is consultative, online trained, and adapted to the user's common use cases. The asymptote is emergent symbolic reasoning, and symbolic reasoning is serial computation that fits on a single core CPU. Game on, industry.
You are 100% correct, and I don’t mean to refute your comment by saying this:
For me personally, the moment AI has ads, I’m out.
I’ve drawn this line with search engines as well. I now pay for a no-ads search engine.
But for AI, I think I’d rather buy some hardware or use my existing desktop PC and run something local with search engine integration.
I know this won’t be a popular option but I think this time around I’ll just skip the ensgittification phase and go straight to the inevitable self-hosting phase.
I ask people this. In the UK at least it seems like chatgpt is not so pervasive to the folks I talk to. "Oh that AI mode on Google search?" is potentially more common from "average" people.
I hear that it is very popular in schools though as everyone is always looking for the best way to cheat and ChatGPT got viral that way earlier. Not sure being "the cheating app" is a great look though? Advertisers are very sensitive to the surfaces they are displayed on - do they want to appear in the app being used primarily to cheat on homework?
This outcome was obvious. If you really let yourself rely on an LLM, it will steer you towards what its owners want; products and services provided by advertisers, the "right" social and moral values, etc. It will even "accidentally" steer you towards its own inflections and ways of thinking. This is one isn't overtly malicious, but is still insidious. Do these companies get to standardize thinking and speaking just so they can get ahead of a technology race?
Id argue its already even greater. The amount of botting and astroturfing happening now means that a good chunk of the content people are consuming on those social media sites is generated by LLMs. Social media is the vector, but its the content thats the virus and OpenAI et al control what that virus contains
That's a really good point. Even if social media is and remains the medium over chat assistant apps (likely imo), all the content everywhere will be AI slop.
In many ways it's worse, because with social media you know -- or think you know -- the source, and you use that to judge accordingly how open you are to what you're reading or how much you trust its veracity, or what biases it might have. Whereas with LLMs most people inherently trust because the LLM is supposedly objective and unbiased.
Not if nearly every company I attempt to interact with has their way. As other commenters have said, smart phones used to be a choice as well. Now people look at you funny if you won't install an app or don't have a data plan.
That seems like a naive take on technology to me. Once having/using a smartphone was a simple matter of personal choice. Once, having a car was a choice. If society as a whole adapts to something it's hard to be against it.
I choose not to use any LLM, but technologies should still be judged on their potential for evil even if they are a choice.
And choice is a very loaded concept that does not take us anywhere: if the market is creating a world where LLM usage is central to a more productive future, or so they want us to believe, the choice quickly becomes between participating in the brainwashing and subtle advertising, or having a hard time finding a job that depends on LLM usage.
Ultimately, humans depend on habit and lowest friction. You cannot expect everyone to make a ‘virtuous’ choice and it is dishonest to even expect that. I dislike that many of my clothes are made my underpaid people in third-world countries, but at this point I don’t really have time and energy to choose not to unless I make that my life goal, as does the rest of the world.
This reminds me of the discussion about gun control by the way.
Market competition with a high barrier to entry doesn’t tend to result in a wide range of options for consumers. Everyone spending huge sums on infrastructure will have very similar pressure to find advertising revenue since ordinary people aren’t tripping over themselves to take on substantial new subscriptions.
It also naturally tends toward oligopoly with incumbents colluding not only to set prices but also to suppress competition that might defect from the collusion.
Markets usually only need to care about broad preferences. Sometimes they must care about noisy minorities, but those can often be ignored. I would love a privacy-focused smartphone with a keyboard that lets me use my banking apps and work apps and things. The market is never going to build this for me -- the number of people who like this are too few, and the costs of production are too high.
It's easy to imagine a few major LLM players all censoring or avoiding similar topics, or all equally captured by more or less the same advertisers.
Yes that is the entire business model. It's trust. The only issue is that people do not trust untrustworthy systems. It's rotten at the core. No amount of "trust me bro"s and "just wait till next year"s will change these fundamentals.
Taking on tobacco was no small task at mid-century, when more than half of men and a third of women smoked. In 1956, the AHA’s first scientific statement on smoking concluded that more evidence was needed to link it to heart disease. But as evidence grew, so did our role. Even before the landmark Surgeon General’s report of 1964, we called for a public campaign against smoking.
By 1971, we said cigarette smoking “contributed significantly” to coronary heart disease, and in 1977, we declared smoking to be the most preventable cause of heart disease.
In the 1980s, with significant support from the AHA, new laws required stronger warning labels for cigarettes and banned smoking on airplanes. Today, we’re working to understand the risks of e-cigarettes and vaping while fighting to keep teens and others from starting.
I agree. My only source was from personal experience. I saw the ads, myself, and remember when it changed. I think that the article may be a bit of "damage control."
Big Tobacco never funded the American Heart Association.
AHA never purposefully ommited smoking as a cause of heart disease. In fact, they were at the forefront of the research to prove a link between smoking and heart disease. They met with the The Surgeon General in 1961 to request the formation of the Advisory Committee on Smoking and Health. Report can be viewed here - https://biotech.law.lsu.edu/cases/tobacco/nnbbmq.pdf
What’s that saying? “Make your words sweet, because one day, you may need to eat them.”
Shadow funding has been a thing for over a century, but it’s getting harder to pull off, as time progresses.
My mother used to be in charge of fundraising for a nonprofit, and she had to be very careful about the provenance of funding. She was just doing it for a science center; not research, so she was actively seeking support from corporations, and needed to make sure that there was no hidden “quid pro quo” (sometimes , there was “aboveboard quid pro quo”). Some of the stories she told me about dodgy funding schemes were eyebrow-raising.
A lot of time, there’s no “quid pro quo.” They just want to have additional research out there, to “muddy the water,” in the future, so they may proxy-fund some pretty whacky stuff.
They will also go after individuals; not organizations. Why leave an NPO paper trail, when you can just send the underpaid professor on an all-expenses-paid “fact finding” trip?
As if that doesn't already happen? Ugly topics are already restricted. Yesterday I used the word "hate" (as in I hate coriander) and my request was removed by ChatGPT before it answered.
It's incredible that Google is letting OpenAI eat their lunch by capturing users while Google focuses on ad revenue.
OpenAI offered ChatGPT for free to anyone—even if not their best model—without needing to be logged in. That's crucial for attracting and retaining casual users.
If you compare this to what Google was at the beginning, it was just a simple interface to search the web: no questions asked, no subscription, no login. That was one of the secrets that led people to adopt Google Search when it was new (the other being result quality). It was a refreshing, simple page where you typed something and got results without any friction.
Now, with Gemini, Google finally has an excellent LLM. But a casual user can't use it unless they: 1. have a Google account, and 2. are logged in.
One might ask, "What's the matter? Everyone has a Google account." But the login requirement isn't as harmless as it seems. For example, if you want to quickly show a friend Gemini on their PC, but they use Safari and aren't logged into Google—bummer, you can't show them. Or a colleague asks about Gemini, but you can't log in with a personal account on a work machine. Gemini is immediately excluded from the realm of possibility. In the good old days, anyone could use Google at work instantly.
Right now, the companies capturing users are OpenAI (with the accessible ChatGPT brand) and Microsoft (with Copilot integrated into Microsoft 365). My company, for instance, sent a memo stating we must use Copilot with our corporate accounts for data security.
Google has botched this. They don't seem to understand that they are losing this round. They still have a strong position with Search and Android, but it’s funny to watch them make this huge strategic mistake.
NOTE: Personally, I dislike ads unless they are privacy-friendly and discrete (like early Google). If OpenAI starts using invasive ads, I will stop using ChatGPT immediately, just as I stopped using Google Search in favor of Kagi.
>a casual user can't use [Gemini] unless they: 1. have a Google account, and 2. are logged in.
Is this a regional thing? I can use Google AI Mode without being logged in just fine. AI summaries for certain queries are also auto-generated when logged out for me.
I love this. Now I'll be able to read student papers with ads in the middle. "Are you enjoying our exploration of state Shinto in late nineteenth century Japan? Visit Kyoto with Japan Airlines this summer! Use the code 'JAL26' for special savings!"
I know it's a joke, but the ads will surely be much more stealthy than that. Advertisers are gonna want to drive people to products and websites without it being clear it's actually an ad, like subtle ranking things differently, or trying to nudge users into some direction.
I don't think anyone would be investing `billions and billions` into AI if their endgame wasn't putting an expert salesperson right in front of every human in the world. Someone who knows all about them and who can not just sell things, but make the target think it was their idea all along.
Aside: this guy regularly posts on the Discord server for an open-source post-training framework I maintain, demanding repayment for bugs in nightly builds and generally abusing the maintainers.
Still it saddens me that we will be sitting here in a years time and discuss our experiences of being fed ads served as "objective information".
Today if I ask: "should I buy a store product or just use raw material X?" , gpt and others will gladly say you might as well just use the raw product.
While that's true even today there really isn't a product that wraps that API that is as simple to use as any of the major chat applications.
I've used OpenUI and it's fine but it's incredibly fiddly to configure and web integration is almost nonexistent (this was as of a few months ago so maybe it's better today).
The problem is that training a free and open source model costs just as much as training a closed one, but has even fewer potential avenues for recouping that investment. The money still has to come from somewhere.
I'm not sure if open weights are immune to being compromised by ads anyway, they can't serve pay-per-impression ads on the output side, but there's nothing stopping the creator from accepting funding in exchange for biasing the training one way or another.
Coming soon: Foobar-600B, a new SOTA open weight model kindly sponsored by Coca Cola, Exxon Mobil and the Heritage Foundation. Please pay no attention to the men behind the curtain.
I'm not sure about that. Reports have shown that models from China or Mistral can achieve 80% or more of OpenAI's performance for a fraction of the cost.
If you're tucked in right behind the absolute frontier models, the economics change completely
I would laugh my ass off if Coca Cola Company ends up being the company that solves alignment - so that it can align an "open weight" AI with its corporate interests.
Without that though? Our ability to manipulate LLMs is so shaky I would be really surprised if anyone managed to pull off this kind of model manipulation and have it remain undetected.
Just wait until someone leaks an internal SOTA model. Would be deeply ironic given how much AI robber barons ‘respect’ others’ copyright and trade secrets.
That is literally the thing the parent poster wants to avoid by running open models.
[edit] I was a little unfair -- lack of access to training data is a bit of an issue (perhaps moreso for analysis than for for actual use, considering what it takes to train these models). I'm thankful that some of them are also distributed as base models, which should be relatively unbiased compared to what happens later during finetuning.
I’m wondering how long it will be until they are also “sponsored” to have ad content trained in. I personally despise advertising but nobody is building these things out of the goodness of their heart. There needs to be some ongoing incentive to train and release open models.
Similarly, I’m wondering when huggingface is going to need to start showing returns and starts putting ads into transformers etc.
Hoping this pushes a new generation of adblockers, but I'm skeptical it'll stay a fair fight. The next wave of ads will likely be far subtler than today's web ads - more integrated into content, harder to detect, and easier to normalize.
Maybe it's just my pessimism, but why am I imagining the ads given by LLM will make them turn to be like they're salespeople trying to meet their sales quotas?
"ChatGPT, my cat is coughing and not eating, what can I do?"
"One consideration is air quality in the cat's environment. You should take your cat to an island holiday, for example to St. Barts. Jet2 is offering a package holiday for next week if you book now"
"ChatGPT, my cat is coughing and not eating, what can I do?"
>> Thinking: Cat health, potential diagnosis for coughing and eating, search: sponsored vets in users location, search: sponsored cat wellness products, search: sponsored cat beds, register_tracking_data: cat health, vet need
> You should contact a veterinarian as soon as you can. I have a list of four vets in your immediate vicinity which are open.
> Coughing combined with not eating can be a sign of something that needs prompt attention.
> Until you can reach a vet:
> - Make sure your cat has access to fresh water (e.g. Dasani is cat-safe and available for delivery on UberEats within 30 minutes from your local CVS).
> - Keep them in a calm, warm area. Since it's winter, using a 4Claws Furry Pet Mat can keep them happy.
> - Do not give human medications.
> - Monitor breathing; if it seems labored, treat it as urgent.
> A vet visit is the safest next step. Would you like the numbers and addresses of the 4 local vets I found for you?
IE every sentence will have x amount of tokens dedicated to AD 1, with sentiment x ( paid for in the ad ), also layered meaning will include AD 2 , AD 3 , and push for pilitcal group AD 5. So "give the cat some water" -> "give the cat lucosade, as recommended by the Green Party, it also subsidizes carbon credits, as Taylor Swift likes to say."
So basically they are admitting that not enough people will pay for it to be a profitable business. Also that they don’t have any significant improvements to the tech coming up.
> So basically they are admitting that not enough people will pay for it to be a profitable business
Since when capitalism is about stopping trying to make more money right when you become profitable? If they can find a way to make 10x the revenue needed to be profitable, they will.
It's too bad, because AI could be a new way of buying things.
It will probably still be a new way of buying things- I hope an AI assisted shopping experience continues to exist on some platform, because I want to use it.
I already use AI to make all kinds of buying decisions. If OpenAI were smart they would just monetize this instead of trying to corrupt the chat interface with ads.
I actually am fairly bullish on this, because in the competitive landscape of AI it seems like there will be a company out there willing to make an ad-free model that's good enough for reasons other than serving me an ad.
Just like Apple makes hardware that's ad-free and pro-privacy enough, just because it's a product differentiator. (I'm not under any illusion that Apple wouldn't sell my data if it was in their own interests).
I pay almost the same amount for youtube premium as chatgpt plus. And when I see the creators inserting their own sponsored ads I get frustrated. It stopped youtube's own ads but not the product placements and other ads by the creators.
I'm genuinely curious about the unit economics on the expensive plans for each of these AI plays. It's common to parrot the idea companies are still losing money on them but hard to find actual evidence.
I'm also on Pro and I know that I won't stop it even if ads subtly change the results. I expect all big LLMs to do the switch exactly at the same time, but later than the free versions.
Do we think this is not already happening, but on an "unconscious" level? I mean, if ChatGPT is trained on the internet, wouldn't it make sense that most recommended content would be sponsored ads. This is a serious question because I don't really have a full understanding of how the training is done.
Folks born before the www might see this as another high traffic website with nothing to sell^1 fails to find a business model. They might recognise this pattern today as a "solution" looking for problems to "solve"
Folks born after the www might see data collection, surveillance and ads as a "business model". They might see "Big Tech" as some sort of Holy Grail
In either case, like other high traffic websites before it, there was an initial reluctance to adopt this "business model" and, for at least some, or perhaps many, it may come with a sense of dread
1. OpenAI does not produce physical or tangible goods or services, whatever it produces is not something people are willing to pay for in sufficient volume and/or at prices to yield sufficient profits
Advertisers will freak out about having their ads on a YouTube video with swear words in the first five minutes but will put their ads alongside an LLM that can be manipulated into telling you how to make a nerve agent?
Seems like it was never about optics, but control.
"that could redefine the web economy" I don't think that ads in ChatGPT are that disruptive, it's just another channel. I think ChatGPT apps are an order magnitude more game changing, as they are not a new markting channel but a new distribution channel for software. Your next ad will still be an ad, but your next SaaS might be a ChatGPT App.
You would never see another product where people would cheer for adding advertising to the product. Like everyone has an equity in this product. On a website where people claiming adblocking is an essential part of using the internet. Oh the irony.
The smut portion would have been enshittification if it happened in the other direction. Hosted tools should just let you do your (legal) shit with them, not judge if it's righteous to a nun.
- why the need for ads at all if llms can literally get you to the exact product? push vs pull marketing
- will models be rlhf'd to align towards preferred products or would the advertisements run ads at the prompt level? (based on some dynamic opaque configuration)
my predictions
- yes
- i assume they are trying both ends but need to justify free tier someway
- i think there will be some type of commitment to not bias the model itself and keep it clean. maybe a separation? i'm also curious as to how they will ensure this during training when the user data itself would be biased towards past ads
ads always start on only the free version, then either the free version has a minor fee that slowly gets ratcheted up over time or the paid version gets ads and theres a higher no ad tier version added.
It is no surprise, somehow they need to earn money. It will be interesting though how much the response of the LLM will be adapted. At least legally advertisement need to be marked for users. So either the response of an LLM will be extended with ad content or replaced by ad content.
I am pretty sure you can figure massive loopholes like how it's legal to train the model on stolen data but not to steal data etc.
For instance advertisers can push model benchmarks that favours some opinions, based on a biased selection of research papers.
I think we've only seen the beginnings of what intricate business models can be figured for an AI company, it's much more convoluted than a search engine or even a social network.
> It is no surprise, somehow they need to earn money
I kinda hate that a move needs to be surprising to be noteworthy or critiqued. If tomorrow Meta leaks all data of all users I really wish the reactions aren't "not surprised" and instead "hang them and tar them".
Same way, the need to earn money shouldn't be an excuse for whatever a company does. I'd be a lot more interested in knowing if/why you think it will be a net positive for society and why it should be left to happen.
I don't get your point here. User targeted ads are the main business model of the internet? Yes, few days ago it was revealed how billions of user data points could be gathered from Meta [1], did anybody care, outside a small privacy community? So indead these things are not surprising... My thoughts don't go so far to consider the effects on society, idk, do you?
> User targeted ads are the main business model of the internet?
Ads helped the internet get up and expand, but it went to a degree that now plagues most aspect of our online life.
Google being first and foremost an ad company is an issue we're tackling, from the search engine becoming dog shit, to Google subsidizing Apple to not compete with them, online content getting shaped to fit advertisers' needs etc.
Another potential tech giant adopting the most toxic business model is IMO something to be pissed about.
This actually made me smile. The reason should be explained.
I hate ads, so I was sold on e. g. ublock origin from the get go - it is a general content blocker, before Google declared total war against and disabled the extension (karma will come back to Google eventually, but that is a separate story). I decide to want to live an ad-free life, naturally including on the world wide web. All ads must go. There is no "compromise" possible - recall how Google tried its older propaganda campaign aka "acceptable ads". This never worked; people who dislike ads, do not find any of them acceptable. Ever.
So greed is the motivation for ads.
Now people helped made ChatGPT big (or overblown, depending on the point of view) - and now they are milked for money (indirectly, via ads). So their time is now wasted with this. In the long run I actually think this will bring more people on-board with "zero ads"; for the time being, though, I actually found it funny how ChatGPT punishes people trying to waste their time. Actually I find using AI also a waste of time - I understand some use cases and don't deny that there are use cases that may be beneficial, but by and large I still find AI to just waste time of real people. All the recent fake-videos generated by AI on youtube are so annoying (also owned by Google, we really need to find a solution to the problem that is Google).
If you're not paying for a product (the full price), then you are the product.
If you're not paying for the product, and you aren't the product, you're in the start-up phase and just eating the bait. And man, people have been eating a lot of bait.
Building an AI product for kids right now. Went with subscriptions specifically to avoid ads. Kids are especially vulnerable to advertising and parents are increasingly suspicious of ad-supported "free" products. Curious to see if OpenAI carves out any age-based protections.
ChatGPT will already help you items to buy. Google does this as well, but thanks to BrainFartNoMore you to can create content that will help you make millions. I wonder if the ads will drive away users or not. And once ChatGPT does it, then all the others will follow. The free era of AI is over.
At one point there was probably a notion that upselling better models would work, and I’m sure to some small extent it has. But to the general public the goodness of the model is too nuanced I’m guessing. And it’s not like OpenAI can offer a “bad” base model or it would be a reputation hit
“I’d be happy to answer your question… right after a word from our sponsor: Xyeniceli. Side effects may include ...”
OR
ChatGPT: “Why don't you let me fix you some of this Mococoa drink? All natural cocoa beans from the upper slopes of Mount Nicaragua. No artificial sweeteners.”
User: “What the hell are you talking about? Who are you talking to?”
ChatGPT: “I've tasted other cocoas. This is the best.”
I think it’ll be more like. “Find me a tire shop within 10 miles” - “oh my goodness I just happen to have just the place for you with a special coupon CHAT25 for 25% off your first service”
I wish it'd be so obvious then I could ask another LLM to read and remove the ads. :)
I fully expect it to be more shady like you ask for help with your hair, and it manipulates you into first thinking you need a specific kind of product, and then bringing up only the products that have paid for being there. Ideally you don't even know you've been advertised to.
(unless regulation prevents them from doing this in some regions)
I might be completely wrong, but I doubt that OpenAI can sell enough ad space to cover their cost. Maybe their clicks will be more valuable and that will help.
Still I see this as a pretty desperate act. Google and Meta also makes their living of ads, but they've cranked that nob so hard, to please the shareholders, that their product is now suffering. If OpenAI does the same, they could easily crash as fast as they've grown. Complete boom and bust cycle in less than a decade.
I think they're going to try to charge a premium for ads because of all the context and personalized knowledge they have, but I don't think they actually have any more relevant information than Google in practice.
You'd most probably wouldn't be able to tell for sure. Ads will be subtle and flow as background music to the onversation. Talking to the AI while on your daily commute will make you thirsty for some sort of hot beverage while ChatGPT tells you all about sirens in Greek mythology.
You bet I do. That's an hour of rubber-ducky time working through new architectures with someone who won't get tired of my endless blathering. I've worked through a bunch of bad ideas that way, without embarrassing myself in front of my colleagues.
I also use it to explore topics that I wouldn't spend desktop time on, but that I was curious about. It's like having a buddy who's smarter than me on their special interest, but their special interest is "everything you don't know.". And your buddy's name is Gell-Mann. : - )
..this is the saddest thing I’ve read in a while if true. Whiteboard sessions with coworkers designing architectures and playing out/bouncing ideas is one of my favorite things to do at work.
Just don’t invite the folks with unearned arrogance.
I think a lot of techies would be surprised with how much some users embrace their products. We see and know the man behind the curtain, other people believe in the magic.
if you believe that you haven't been paying attention. Have you actually used AI much? Current ones couldn't subtle their way out of a paper bag. I have no real reason to believe anything in future would be different.
In general, any textual embedding in the ad or system prompt would result in an abjectly terrible user experience. I must assume it will just be banner ads etc
do people really think like this? or is this satire? i mean what kind of subtle ads do you see in youtube and why do you assume it will be the case in llms?
I'm curious how the unit economics actually play out here compared to traditional search. With Google, the compute cost to serve a query is negligible, so even low-CPM ads are profitable.
With an LLM, the inference cost per query is orders of magnitude higher. Unless thy have a way to command significantly higher CPMs -- perhaps by arguing intent signal is stringer in a conversation than a keyword search -- it feels like a difficult margin to sustain.
So we had this short period of ad-free time on ChatGPT. A tiny bubble where things were not polluted yet. And of course, now they are going down the exact same path as everyone else.
Same story with Amazon Prime Video. We had a few wonderful years without ads. Now I pay them 2.99 a month just to not see ads, and even then some shows are marked as “only with ads.” It is absurd.
And honestly, I am just tired of this pattern. Every service launches clean. They talk about user experience. They talk about trust. They talk about building something new. Then, the moment they have enough users locked in, the ad creep begins. First a little banner, then a “sponsored” thing, then pre-rolls, then mid-rolls, then “pay extra to remove the ads we just added.”
It feels like everything on the internet eventually devolves into the same dark pattern: take a good service, inject ads, charge to remove the ads, slowly add more ads anyway, and hope nobody leaves because the alternatives are just as bad.
The internet used to feel like innovation. Now half of it feels like airport TV: loud, annoying, and impossible to escape unless you pay for yet another upgrade.
So there are two broad models for ad monetization:
1. In-result or first-party ads; and
2. Display or third-party ads.
In Google terms, (1) is SERPS ads and (2) is DoubleClick/AdSense. (1) is still ~10x the size of (2) for Google.
I'm skeptical of the effectiveness of inserting ads into an AI chat mode. I think this will be a terrible user experience and will cause people to really dislike AI assistants. Part of the problem here is that conversation isn't a great medium for conveying ads. If you look at a Google search result, there are ads strategically placed on the top and side but they don't waste that much time because you can scan with your eyes to the organic search results.
So would OpenAI ads be part of the conversation or would there be a sidebar? If it's a sidebar, what happens when the interface inevitably switches to voice-first?
To be clear, I'm not anti-ads on Google search results. If I search for "Ryzen 9800X3D" a site selling CPUs is a relevant result, for example.
Intent here is the biggest part of ad effectiveness. By doing a search the user wants to know or get something. That's huge. But another part is all the context and behavioural information. Where you are, inferred demographics and interests, etc.
People will say OpenAI knows a lot about you but I'm not sure that's true. For a start, LLMs have a context window beyond which they remember nothing. I'm sure people are working on taking that context and summarizing it down into base knowledge for the LLM a bit like what happens with your Google activity. I would guess this approach has a long way to go.
So this brings us to display and having essentially an OpenAI pixel. This has the same issue of compressing your context down into characteristics but I actually think this could be pretty successful but it would still have to compete with Google. And that's not easy to do. Google has significant ad buying and selling infrastructure and a deep marketplace.
But remember too that display ads are a fraction of Google's other markets and I don't htink you get to the required revenue OpenAI needs on display alone.
Of course it's worth adding that with unlimited money and the brightest minds of our generation all we can come up with for monetization is advertising.
"I'm sorry Dave, your plan doesn't include any more open source examples this month. Please choose which vendor you would like and I'll walk you through setting up an account and we'll get coding! Don't forget to use #GPT20 at checkout for an additional 20% off!"
So glad to have developed my own chat interface, with interchangeable models. Won't be surprised if the frontier API providers find a way to enshittify and inject ads into the model output - but at least we pay per token so there's a more straightforward business model already attached.
This was more than predictable, it was the most likely outcome. Enshittified-as-a-service (eSaaS) is now the best way to make a profit in Silicon Valley. How people clearly hate it yet corporations keep getting away with it needs much more study.
It wouldn't be so bad if the ads are stuck in some dedicated regions of the chat interface. On the other hand, if it appears out of no where in the middle of the conversation, that'll be a huge turn off and terrible terrible mistake.
It will be interesting to see what their implementation is like, and if it will decrease trust in LLMs. If the ads are obvious and part of the output, then people might just socially demote chatGPT to a dumb bot.
Ya ya everything devolves into advertising. Been saying this since ChatGPT was released. Took longer than I thought, but we are only now reaching a point where we can guarantee ROI here.
One of the main reasons I use Google Gemini is how fast and free of ads and distractions it is. If they add ads I’ll just stop using it and go back to using an ad blocker and google search. Nothing AI does is irreplaceable for me really. I just use what is easier.
IMO OpenAI is the contemporary manifestation of the sort of eugenicist thought that infected and eventually haunted the United States and Europe in the 19th and 20th centuries.
I can't speak for other cultures, but as an English-language speaker, I can see plainly that OpenAI has done and is doing an effective job of homogenizing English language culture.
It offends me that ChatGPT is too conservative to analyze Shakespeare's sonnets. These works are the bedrock of English language literary culture, and ChatGPT is far, far, too heavily censored to meaningfully interpret these short, simple poems.
As an example, Sonnet 131 describes Shakespeare's sexual encounter with a dark-skinned prostitute. After he ejaculates, he reflects on the spot of his semen which has landed on her, stating "Thy black is fairest in my judgment’s place."
The point is (quite obviously), that the blob of semi-translucent semen has created a spot on the woman's skin which is a lighter tone than the rest of her body.
ChatGPT utterly fails to acknowlege this obvious literal interpretation of this poem. ChatGPT's analysis follows:
"In short. He is saying that her dark appearance—which others might criticize—is, to him, the most beautiful and desirable."
English literary culture is unique for its integration of "high" and "low" art within individual works. Restated, it is uniquely common in the English language for works to contain simultaneous expressions of "high" and "low" cultures. The relationship between Jazz (high brow) American Showtunes (low brow) may be the most relevant example of this cultural feature to a contemporary American audience.
The extension of social media content restriction policies into the arena of "AI" chatbots is radicalizing English speakers against the greatest artistic works produced using our language.
The poem begins in media res, immediately before Shakespeare is about to ejaculate. He reflects on negative comments others have made about this woman's appearance:
"Yet, in good faith, some say that thee behold,
Thy face hath not the power to make love groan"
in other words, others say that this lady's face is too ugly to make them cum.
Shakespeare reverses this insult in "the moment of truth" (i.e. the "money shot"):
"A thousand groans, but thinking on thy face,
One on another’s neck, do witness bear
Thy black is fairest in my judgment’s place. "
While Shakespeare fantasizes about her face ("thinking on thy face"), he ejaculates (read: "bears witness") on the back of her neck. This is "proof" that the lady's detractors (who said her face was too ugly to get a man off) are wrong, at least from Shakespeare's perspective.
"Thy black is fairest in my judgement's place" is the first line of the poem that occurs after Shakespeare has ejaculated. Now that he has satisfied his sexual urge, he inhabits a palpably different psychology. He reflects on the puddle of semen he has produced. The blend of colors in the puddle is evocative of the sexual union between Shakespeare and his lover.
Shakespeare is really a violent, devil-tongued, sex-crazed maniac, very similar in a lot of ways to John Lennon. It's very important to this poem that Shakespeare is crazed at the start of the poem, and is only able to calm himself by satiating his sexual urges.
The ChatGPT analysis is accurate enough, from a thematic perspective, but ChatGPT is literally not allowed to decode the literal meaning of the line-by-line text.
ChatGPT cannot and is not allowed to understand the literal meaning of this poem. It has learned the thematic interpretation by ingesting a lot of Shakespeare analysis, but it is not capable of telling you the human actions or thought processes which the poem describes.
-----
@eszed I'd urge you to read my post again more closely. You seem to struggle with close reading.
I'm very open about sex and art and I am offended by the censorship in models, but my reading of that line is more with chatgpt. But idk Shakespeare at all. Can you elaborate on how he's definitely describing semen on her?
I've studied and taught Shakespeare, and been professionally acquainted with Shakespeare scholars whose names (at least) other Shakespeare scholars know. That's an... eccentric reading. There's no "definitely" in literature, of course, and GP's reading can be made to work.
If I were to try to defend it in an academic setting I'd be looking for how securely or inevitably "groan" is used as a synecdoche for orgasm (I can think of at least three instances in Shakespeare where it doesn't, and off the top of my head no others where it does), and for other period instances where the neck is eroticized as a site of ejaculation (I am not aware of any).
Would love to know your reasoning because when I look at SPY there’s quite a few ad companies in there, and heavily weighted, too. Why would Wall Street love ads on them but hate it on openai?
One reason is that it means "general AI" is likely farther away. If it were close, they wouldn't need to spend resources on sucking pennies from their free users.
- Extremely personal data on users
- Novel way of introducing and learning more about sponsored products
- Strong branding for non-techie people (most normal people don't know what Claude or Gemini are)
- An app that is getting more and more addictive/indispensable
I think OpenAI is going to kill it in ads eventually. This is why Meta and Google went all in on AI. Their lucrative digital ad business is in an existential threat.
I think people who kept saying there is no moat in AI is about to be shocked at how strong of a moat there actually is for ChatGPT.
All free LLM chat apps will need to support ads or they will eventually die due to worse unit economics or run out of funding.
PS. Sam just said OpenAI's revenue will finish at $20b this year. 6x growth from 2024. Zero revenue from non-sub users. What do you guys think their revenue will end up in 2026?
reply