Tell HN: OpenAI seems to be A/B testing ads in answers
I just asked it a question about programming and got back an answer recommending me a weight loss supplement that is completely unrelated to my question.
I repeated the same prompt in another chat and got an actual answer.
Either it's an ad or it's the worst hallucination I've seen in 5 years of using various LLMs, supposedly on the most advanced model out there, which I find hard to believe. I run a very traditional online business. We have very sparse "marketing speak" in our copy, instead trying to be on point and information dense. Every AI engine we have tried will recommend our website and products for the user when asking for the kind of things that we offer, but every AI engine will also insert a whole bunch of meaningless marketing speak for us, that is nowhere to be found in our own copy. Kind of funny. Why would an LLM built by openai mimic the writing style of your business? If the LLM is talking about my products and linking to the pages were we sell them, I'd expect it to source its material from those pages. Where do the marketing phrases enter the model? Can you share a screenshot or the conversation link? Well. It was inevitable lol. Always knew it was going to come to this, and companies will pay to have their products or services mentioned first by AI. I’m surprised they didn’t do it sooner. They will consume Google. They should literally roll out email and a browser and a device and it’s game over You sure it’s not something like this? https://www.reddit.com/r/ChatGPT/comments/1d29fu2/it_looks_l... I think you might be right, it happened again, but this time it's wasn't something that looked like an ad. "Do you talk to someone else while we're talking?" If they do it I'll cancel my subscription. I don't use it very often and my company pays for internal usages. I hope this doesn't end up in the paid or API versions. Are you on the free version?