> you can let Copilot see all your open tabs so it can understand the full context of what you’re exploring online
Isn't that going to poison the context with low quality, SEO blog spam? How long is it going to be before content mills start filling the web with defensible lies?
For example, "exploding cell phone charger X is built with a high quality transformer." The AI will mindlessly regurgitate that without considering the nuance of source. When I evaluate that stuff the source plays a major role in the weight I give to the statement.
None of the AI I've seen does a proper job of considering the source and I'm guessing it's because they can't attribute sources due to the massive IP theft that was required to build the systems.
Google already has page score for every website out there. What makes you think that their LLM can't be made to use it? That statement sounds a lot like "AI can't generate human fingers, it will never be able to generate realistic images with humans in them" or "it can't do X so it will never be able to code Y". Shortsighted.
Google search results are much worse than they used to be. And how would an LLM extrapolate good outputs from garbage inputs?