As model performance converges, it becomes the strongest moat. Why go to Claude for a marginally better model when you have the ChatGPT app downloaded and all your chat history there.
Do you ever look at your chat history? why? how far back?
I actually pre-emptively deleted ChatGPT and my account recently as I suspect that they're going to start aggressively putting ads and user tracking into the site and apps to build revenue. I also bet that if they do go through with putting ads into the app that daily user numbers will drop sharply - one of ChatGPT's biggest draws is its clean, no-nonsense UX. There are plenty of competitors that are as good as o1 so I have lots of choices to jump ship to.
The day LLM responses start containing product placements is not far now.
And some of them will be from poisoned data, not just an explicit prompt by the site-owner.
And some of them will be from poisoned data, not just an explicit prompt by the site-owner. A whole new form of spam--excuse me--"AI Engine Optimization."
True. ChatGPT has already won me over. I stopped even testing anything else.