Leak confirms OpenAI is preparing ads on ChatGPT for public roll out

Putting ads into ChatGPT could be risky: the system accumulates extensive personal information, the resulting hyper-targeted ads might become persuasive enough to influence what people click on.

I really hope someone can come up with fast local inference hardware good enough to stop using cloud providers.

On one hand, true, my mom still uses Yahoo. But email has a strong network effect - you need to update everyone & every service who has your @yahoo address. Switching does happen there are no network effects. Nobody uses Mapquest or Ask Jeeves.

What might these ads actually look like in practice? It's not immediately obvious to me, if I'm being honest.

I don't think it is impossible, but suggesting you can go from zero to competing with Google and Facebook with a few dozen people is wildly underestimating the requirements. Yes a few dozen engineers could probably build and support the infrastructure to run it, but that is not the hard part. The hard part is the sales, is the compliance, the account management, is the billing system and inevitable snafus, is the first second and third line of tech support, the wining-and-dining, the networking, the AdChoices stuff, the sales calls and pitches, the conversion tracking, the legal arguments and contracts, the gearing up for the clients' big holiday campaigns, the management and general feeding and watering of those hundreds and thousands of people etc as a business, and then also building and supporting the self-serve platform for all the advertisers with credit cards but who are too small to get to talk to real sales people (and how do you support and service those clients too? Billing, reporting, charge backs, stolen cards, fraud, bad ads, more fraud etc - AI and automation can do some of this yes, but it's not like the market leaders aren't already doing this)

When you're working in ads, you don't have 1 person that looks after multiple huge clients.

If you have a Coke or a Nike or a McDonalds or an Apple etc spending 10s/100s of millions in ad spend on your platform, you have a dedicated whole team of 3 or 4 or 5 or more people (sales and dedicated tech support, plus managers etc) per client who exist solely to make it easy for that client to run ads on your platform, and make sure they're happy and getting results. So just those 4 clients you are probably looking at 20+ people just in sales/after-sales supporting 100+mil of ad spend, and that is before you need to support agencies like WPP et al that are often teams of 10 to 15 or more. And if a client doesn't think they're getting the results they need or the treatment they need, they'll take their ad spend elsewhere - this is why you have multiple sales people and hands-on tech people swarming the big spenders to keep them happy and keep them spending. They won't be happy talking to an AI chat bot when their Thanksgiving campaign has gone offline and no one knows why - they'll want their dedicated person to help them get back online ASAP.

It is a huge undertaking to pivot to become a large-scale ad network. Not saying it is impossible, but it is not quick nor easy by any stretch, and should not be underestimated.

Genuinely their best bet is to start by selling ad slots via Google (and you can bet there will be a team of 10+ sales and tech support ready to support them exclusively within a few days if they do) while they build their own capability (if at all). Google ads will have better tracking and targeting and so better conversions than Open ai could do themselves due to the network effect of Google's existing online properties (e.g. YouTube, Search, Play etc), partner ad network, browser, and mobile dominance (despite what people on HN think, the "normal" people online do click on ads and retargetting (i.e. ads following you around) does get results for advertisers. This is why Google and Facebook are printing money).