Remember the internet before algorithmic ads and cross site tracking?
We will remember this moment of LLM usage for the years to come as we are irreparably spun by advertisers in our most intimate and private 1:1 conversations with these AIs
The writing has unfortunately been on the wall for this, especially for free users. They want you to make choices that are in their economic interests.
The biggest tell for me lately is that if you ask ChatGPT about products or even specific items on Amazon, it will only return links to companies that partner with OpenAI. None of the companies they currently partner (affiliate link basically) with are ones I prefer to buy from. It has made ChatGPT way less useful for this kind of research now. I certainly have reasons to distrust all their shopping links.
>Mission alignment: Our mission is to ensure AGI benefits all of humanity; our pursuit of advertising is always in support of that mission and making AI more accessible.
Advertising and marketing kills humanity, these should be among the first industries that AI eliminates; kinda getting mixed messages here. You'd think the tech that supposedly is going to make money and all work irrelevant could figure out a way to make money without resorting to being yet another mechanism to deliver ads.
| Long-term value: We do not optimize for time spent in ChatGPT. We prioritize user trust and user experience over revenue.
The unspoken part -- This holds true so long as revenue is at least equal to costs, and speaks nothing about whether user trust and user experience is optimized over profit.
I’ve been paying for Google Workspace for my custom domain for years basically just so I can use Gmail. For just $7 more dollars a month, I upgraded my plan to access Gemini Pro, which has guaranteed enterprise-grade privacy controls. I think this is currently the best value platform for anyone who values their privacy for LLMs. If Apple and the DoD trust Google’s internal controls, I do too.
they have been doomed for a while, it is just a matter of time, but honestly i like them better than the claude provider, if they can make openai profitable, that would be good for all of us, we don't want a world where gemini is the only winner or the chinese take over
Is there any hope that they wont let any of the ads logic into the rl / pre-training? I'd like for my paid ChatGPT model to offer an unbiased source of truth on what the best products are.
"In the coming weeks, we’re also planning to start testing ads in the U.S. for the free and Go tiers, so more people can benefit from our tools with fewer usage limits or without having to pay."
When ads were introduced into Google, it objectively made the search product worse. You no longer saw what they thought would be the most relevant result. You literally saw Google's most profitable (it is an auction) result.
If we apply that kind of thinking to chat LLMs, it means instead of getting the most relevant tokens, you'll be shown the most profitable tokens. Maybe the most relevant tokens will be below the fold (like unpaid search results) now.
Ironic to show ads only to your users with the least spending power. Me thinks this will not last. Once the ads get “good enough” they will be everywhere.
Oh, this is so benign and appropriate. This will never escalate, and OpenAI is governed by strict privacy laws and audited by the public so we can trust they won't ever change their policy or have bias injected into their models.
Curious how the economics are going to go. ChatGPT has close to a billion MAU and they're losing money even with subscriptions. Meanwhile Google with Gemini is catching up (although I've seen reports that Gemini is also testing ads in their chats, especially without disclosing it's an ad, by linking to Google Shopping [0]). The Google example might show how it might appear:
> Since you are context-switching between multiple projects, you cannot afford to work from a kitchen table with poor ergonomics. These tools make 2-week "sprints" viable:
> Reliable Productivity A portable second monitor is essential for keeping your Slack/Jira open on one side while you code on the other.
> The ASUS ZenScreen is the gold standard for nomads. It's ultra-thin and connects via a single USB-C, meaning you can set up a "pro office" in any Airbnb in under 60 seconds.
> Redundant Connectivity Don't trust Airbnb Wi-Fi for your "Impact Doc" delivery or lead syncs.
> The Netgear Nighthawk M6 provides a dedicated 5G connection. It’s your insurance policy against a bad router in a New Orleans rental house.
> Audio Privacy In urban areas like Chicago or New Orleans, street noise is a constant risk during meetings.
> The Sony WH-1000XM5 headphones have the best background noise cancellation for your microphone. Your manager will hear your voice clearly, even if there's a siren outside your window.
Incidentally, I just saw on Show HN an AI SEO tool [1], wonder if OpenAI will also include similar features in their ad platform. Maybe we'll just type in our questions and it'll spit out stats and graphs itself, now that is more likely.
25 comments
[ 2.7 ms ] story [ 43.6 ms ] threadWe will remember this moment of LLM usage for the years to come as we are irreparably spun by advertisers in our most intimate and private 1:1 conversations with these AIs
Then you move to the paid plan, then they move ads to the paid plan, then you move to the premium-extra plan, etc.
The biggest tell for me lately is that if you ask ChatGPT about products or even specific items on Amazon, it will only return links to companies that partner with OpenAI. None of the companies they currently partner (affiliate link basically) with are ones I prefer to buy from. It has made ChatGPT way less useful for this kind of research now. I certainly have reasons to distrust all their shopping links.
Advertising and marketing kills humanity, these should be among the first industries that AI eliminates; kinda getting mixed messages here. You'd think the tech that supposedly is going to make money and all work irrelevant could figure out a way to make money without resorting to being yet another mechanism to deliver ads.
The unspoken part -- This holds true so long as revenue is at least equal to costs, and speaks nothing about whether user trust and user experience is optimized over profit.
Threads introduced ads and almost no-one cared. This will be the exact same result.
If you don't want the ads, then pay $8 for ChatGPT Go.
"In the coming weeks, we’re also planning to start testing ads in the U.S. for the free and Go tiers, so more people can benefit from our tools with fewer usage limits or without having to pay."
If we apply that kind of thinking to chat LLMs, it means instead of getting the most relevant tokens, you'll be shown the most profitable tokens. Maybe the most relevant tokens will be below the fold (like unpaid search results) now.
Expect all other LLM vendors to follow soon. If they don't, they will lose.
I don't see why I should believe this.
> Since you are context-switching between multiple projects, you cannot afford to work from a kitchen table with poor ergonomics. These tools make 2-week "sprints" viable:
> Reliable Productivity A portable second monitor is essential for keeping your Slack/Jira open on one side while you code on the other.
> The ASUS ZenScreen is the gold standard for nomads. It's ultra-thin and connects via a single USB-C, meaning you can set up a "pro office" in any Airbnb in under 60 seconds.
> Redundant Connectivity Don't trust Airbnb Wi-Fi for your "Impact Doc" delivery or lead syncs.
> The Netgear Nighthawk M6 provides a dedicated 5G connection. It’s your insurance policy against a bad router in a New Orleans rental house.
> Audio Privacy In urban areas like Chicago or New Orleans, street noise is a constant risk during meetings.
> The Sony WH-1000XM5 headphones have the best background noise cancellation for your microphone. Your manager will hear your voice clearly, even if there's a siren outside your window.
Incidentally, I just saw on Show HN an AI SEO tool [1], wonder if OpenAI will also include similar features in their ad platform. Maybe we'll just type in our questions and it'll spit out stats and graphs itself, now that is more likely.
[0] https://news.ycombinator.com/item?id=46533480
[1] https://news.ycombinator.com/item?id=46642490