I share many of your ideas, and I think the best solution would be:
1. a pure data API web (like the original semantic-web idea)
2. open-source browsers which can query for information using on-device LLMs and display it to the user in any UI way they want.
I think 1. will happen, since all search engines will use AI results, with no click through to the original data-owner (website). So there is no more financial incentive to keep a UI website. The question is if the "data API web" will be decentralised or under the control of a few big players that already mined the web.
2. will hopefully happen if on-device models become more capable, the question is by then whether most people are already defaulted to AI browsers from big tech (since they have the money to burn-cash using cloud LLM services to capture market share before on-device LLMs are good enough). The only way to prevent this is user-education and mistrust verus Big Tech, which is what already befell Microsoft's Recall (besides a terrible security architecure).
1. a pure data API web (like the original semantic-web idea)
2. open-source browsers which can query for information using on-device LLMs and display it to the user in any UI way they want.
I think 1. will happen, since all search engines will use AI results, with no click through to the original data-owner (website). So there is no more financial incentive to keep a UI website. The question is if the "data API web" will be decentralised or under the control of a few big players that already mined the web.
2. will hopefully happen if on-device models become more capable, the question is by then whether most people are already defaulted to AI browsers from big tech (since they have the money to burn-cash using cloud LLM services to capture market share before on-device LLMs are good enough). The only way to prevent this is user-education and mistrust verus Big Tech, which is what already befell Microsoft's Recall (besides a terrible security architecure).