In the last few months I noticed that Firefox was eating all of my RAM. Then I heard the community has been complaining about AI features added to FF. I think Firefox has some great ideas for adding value with some AI features. Also remember it's early days. We need to experiment to find the best ways to use AI. The problem with how AI was implemented in Firefox is there was a complete lack of communication. Firefox should have popped up a dialogue saying "New AI features available in FF" "Do you want to enable a built-in model, the resource requirement is 3GB RAM" or whatever. Instead of just eating RAM without telling anybody. I looked at the AI controls that you've added, and it's a step in the right direction. The thing is I would NEVER want FF to load it's own model into RAM for me. I'd much rather tell FF to use my API key and URL of my local self-hosted Ollama or LM-Studio instance, where the model is held in VRAM on my GPU, not in my system RAM. Regardless of that, I might want to just give my enterprise API key, because I've configured my enterprise API accounts such that the provider may not train on my data. The point is I want control. Not only which features to enable. What model to run. Local vs remote. For each feature. These are essential options. The AI features should be open source distinct addons, on Github. So that the community can fork and improve your AI features. Then you will get rapid innovation and potentially much faster adoption of Firefox. Remember the ethos of Firefox. Firefox users are typically smarter, discerning and know what they want. Give us more controls. I want to control how the features work, we should be able to set blacklists so the link summary feature doesn't run in places where we don't want summaries.
... View more