While streaming from my shared screen or from my camera, it seems like the tool and the model is overly focusing its knowledge on what’s being streamed rather than from its general knowledge. For example:
Me: who is Lebron James?
AI: I don’t have that information from your screen
Me: ignore my screen, who is Lebron James?
AI: (a brief sentence about Lebron James)
What’s worse is that let’s say I Googled up Lebron James and the Google results shows Gemini a number of basic information about him, the AI will then refuse to answer other well known facts about Lebron that should be part of any language model, such as the first team he’s played first.
This creates a very limiting experience. Is this behavior intentional? If not, what should I do instead?