> Just think about how much more effective advertisements are going to be when LLMs start to tell you what to buy based on which advertiser gave the most money to the company.

Optimistic view: maybe product quality becomes an actually good metric again as the LLM will care about giving good products.

Yea, I know, I said it's an optimistic view.

What does it mean for the language model to "care" about something?

How would that matter against the operator selling advertisers the right to instruct it about what the relevant facts are?

I think it might be like when Grok was programmed to talk about white genocide and to support Musk's views. It always shoehorned that stuff in but when you asked about it it readily explained that it seemed like disinformation and openly admitted that Musk had a history of using his business to exert political sway.

It's maybe not really "caring" but they are harder to cajole than just "advertise this for us."

For now anyways. There’s a lot of effort being placed into putting up guardrails to make the model respond based on instructions and not deviate. I remember the crazy agents.md files that came out from I believe Anthropic with repeated instructions on how to respond. Clearly it’s a pain point they want to fix.

Once that is resolved then guiding the model to only recommend or mention specific brands will flow right in.

Golden Gate Claude says they know how to do that already.

https://www.anthropic.com/news/golden-gate-claude

large language models don't "care" about anything, but the humans operating openai definitely care a lot about you making them affiliate marketing money

Optimistic view #1: we'll have AI butlers between the pane of glass to filter all ads and negativity.

Optimistic view #2: there is no moat, and AI is "P=NP". Everything can be disrupted.

Has a tech company ever taken 10s or 100s of billions of dollars from investors and not tried to a optimize revenue at the expense of users? Maybe it's happened but I literally can't think of a single one.

Given that the people and companies funding the current AI hype so heavily overlap with the same people who created the current crop of unpleasant money printing machines I have zero faith this time will be different.

1 Trillion US dollars?

1 trillion dollars is justified because people use chatGPT instead of google sometimes?

Yes. Google Search on its own generates about $200b/y, so capturing Google Search's market would be worth $1t based on 5x multiplier.

GPT is more valuable than search because GPT has more control over the content than Search has.

Why is a less reliable service more valuable?

It doesnt matter if its realiable.

Google search won’t exist in the medium term. Why use a list of static links you have to look through manually if you can just ask AI what the answer is? Ai tools like chatgpt are what Google wanted search to be in the first place.

Because you cannot trust the answers AI gives. It presents hallucinated answers with the same confidence as true answers (e.g. see https://news.ycombinator.com/item?id=45322413 )

for now

Aren't blogspam/link farms the equivalent in traditional search? It's not like Google gives 100% accurate links today.

exactly. AI is inherently more useful in its form.

google search engine is the single most profitable product in the history of civilization

In terms of profit given to its creators, “money” has to be number one.

ChatGPT will have access to a tool that uses real-time bidding to determine what product it should instruct the LLM to shill. It's the same shit as Google but with an LLM which people want to use more than Google.

> Just think about how much more effective advertisements are going to be when LLMs start to tell you what to buy based on which advertiser gave the most money to the company.

This has been the selling point of ML based recommendation systems as well. This story from 2012: https://www.forbes.com/sites/kashmirhill/2012/02/16/how-targ...

But can we really say that advertisements are more effective today?

From what little I know about SEO it seems nowadays high intent keywords are more important than ever. LLMs might not do any better than Google because without the intent to purchase pushing ads are just going to rack up impression costs.

> when LLMs start to tell you what to buy based on which advertiser gave the most money to the company.

isn't that quite difficult to do consistently? I'd imagine it would be relatively easy to take the same LLM and get it to shit talk the product whose owners had paid the AI corp to shill. That doesn't seem particularly ideal.

I mean I think Ads will be about as effective as they are now. People need to actually buy more and if you fill LLMs with ad generation well the results of results will just get shitty the same way googles search results had. Its not a Trillion dollar return + 20% like you'd want out of that investment

> Just you switching away from Google is already justifying 1T infrastructure spend.

How? OpenAI are LOSING money on every query. Beating Google by losing money isn't really beating Google.

How do we know this?

Many of the companies (including OpenAI) have even claimed the opposite. Inference is profitable; it's R&D and training that's not.

It's not reasonable to claim inference is profitable when they've also never released those numbers. Also the price they charge for inference is not indicative of the price they're paying to provide inference. Also, at least in openAI's case, they are getting a fantastic deal on compute from Microsoft, so even if the price they charge is reflective of the price they pay, it's still not reflective of a market rate.

DeepSeek on GPUs is like 5x cheaper then GPT

And TPUs are like 5x cheaper then GPUs, per token

Inference is very much profitable

You can do most anything profitability if you ignore the vast majority of your input costs.

OpenAI hasn't released their training cost numbers but DeepSeek has, and there's dozens of companies offering inference hosting of open weight models for the very large models that keep up with OpenAI and Anthropic, so we can see what market rates are shaking out to be for companies that have even less economies of scale. You can also make some extrapolations from AWS Bedrock pricing and can also investigate inference costs yourself on local hardware. Then look at quality measures of quantizations that hosting providers do and you get a feel for what hosting providers are doing to manage costs.

We can't pinpoint the exact dollar amount OpenAI categorically spends but we can make a lot of reasonable and safe guesses, and all signs points to inference hosting being a profitable venture by itself, with training profitability being less certain or being a pursuit of a winner-takes-all strategy.

Statistically this is obvious. Most people use the free tier. Their total losses are enormous and their revenue is not great.

No, it’s not obvious. You can’t do this calculation without having numbers, and they need to come from somewhere.

Sam has claimed that they are profitable on inference. Maybe he is lying but I don't think speaking so absolutely about them losing money on that is something you can throw around so matter of fact. They lose money because they dump an enormous amount of money on R&D.