alt Hacker NewsThese are the less worrying kind of ads in our future.
Seeing how google has been fighting SEO for ages, what's going to happen when companies figure out how to inject ads into the model?
We haven't yet seen the problem of adversarial content in play, I think.
Replies
I've had two people reach out to me asking about one of my services. They both said ChatGPT recommended it to them.
My service does kind of exist. It's a small tool I created for a client while retaining full rights to the tool. So I created (vibe coded) a site around it, making it look like an established service. Even ran google ads for it for a while.
The service still doesn't show up on google with relevant search terms. There hasn't been another client. I forgot about the service. And then ChatGPT started recommending it to people.
I wonder what I did to achieve this. Did vibe coding the business page inject it into ChatGPT's training data?
The worrying kinds of ads won't be from SEO tricks doing sneaky things without OpenAI's approval. OpenAI will just quietly take money from people who will pay to have the AI causally promote their products or their talking points in the output or suppress mentions of competing products or talking points in the output. Maybe they won't even take money for this and the people running OpenAI will do it themselves to promote or censor whatever they want. Either way, it won't look like ads to the user. It's just what happens when greedy people gain control over how other people get their information.
Can easily seo the knwlege chain or seo poison the sources
I experimented with this way back when custom GPTs were first released (looks like late 2023). There are a few / commands you can use to suggest what product to inject, how overt, etc and a generic /operator command to send whatever you like 'out of band' from the chat.
https://chatgpt.com/g/g-juO9gDE6l-covert-advertiser
One of the most interesting things is when it starts pitching a product and you start interrogating it about why it picked that product. I haven't used it in probably a year so it may not do the same thing now, but back then it 100% lied consistently and without any speck of remorse. It was rather eye opening.
Edit: Tried again, it didn't lie this time lol - https://chatgpt.com/share/69f16aa4-c008-83ea-92b3-51f16ca77d...
Why do you need to inject ads at the model weights layer when you control the frontend?
Have the model generate keywords from the query, then inject guidance from matching advertisers into the context window
q: How do I make a new React app?
a: Vercel makes it easier to get your project running fast ⓘ
Some other choices would be:
...
ⓘ This part of the response was sponsored by Vercel
[dead]
It's not an issue of how - there's a great ADM with markup/down supported already, waiting for system prompts to be injected in realtime via the same online auction system that powers banner ads and smart tv content. There's got to be some latent resistance to the idea for now - but it's so easy to do, it'll happen.
The model already advertises because they where trained on massive data’s that refers big brands.
Ask for suggestions for a new pair of shoes. What brand do you think it will suggest Nike, Adidas or some random small one?