[erkt]

I just asked O3 for a software stack for deploying AI in a local application and it recommended llama over OpenAI API.

It has always been easy to imagine how advertising could destroy the integrity of LLM's. I can guarantee that there will be companies unable to resist the temporary cash flows from it. Those models will destroy their reputation in no time.

[xyzal]

It's happening, though not with ads yet

https://www.washingtonpost.com/technology/2025/04/17/llm-poi...

[cogman10]

I'm an AI pessimist, yet I don't see this happening (at least not without some major advancements in how LLMs work).

One major problem is the payment mechanism. The nature of LLMs means you just can't really know or force it to spit out ad garbage in a predictable manor. That'll make it really tricky for an advertiser to want to invest in your LLM advertising (beyond being able to sell the fact that they are an AI ad service).

Another is going to be regulations. How can you be sure to properly highlight "sponsored" content in the middle of an AI hallucination? These LLM companies run a very real risk of running a fowl of FTC rules.

[handfuloflight]

> The nature of LLMs means you just can't really know or force it to spit out ad garbage in a predictable manor.

You certainly can with middleware on inference.

[rurp]

It matters a lot how much of the market they capture before then though. Oracle and Google are two companies that have spent years torching their reputation but they are still ubiquitous and wildly profitable.

[JoshTko]

My bet is that free versions of models will become sponsor aligned.