Google on Wednesday began inviting Gemini users to let its chatbot read their Gmail, Photos, Search history, and YouTube data in exchange for possibly more personalized responses.
Josh Woodward, VP of Google Labs, Gemini and AI Studio, announced the beta availability of Personal Intelligence in the US. Access will roll out over the next week to US-based Google AI Pro and AI Ultra subscribers.
Even that is just confusing. I sometimes use Perplexity (because Pro comes with my bank account - neobanks have zero focus). And by default it remembers things you say. So when I ask a question sometimes it will randomly decide to bring in something else I asked about before. E.g. I sometimes use it to look up programming related stuff, and then when I ask something else it will randomly research whatever language it thinks I like now in that context too and do things like suggest an anime based on my recent interest in Rust for no good reason.
That’s true, these supposedly intelligent systems are really stupid about this stuff still. Especially with limited room to store that additional ever growing context about you.
I wasn’t accounting for quality, and it’s bad right now for this. And I’m skeptical it will get better. The models need to be tactful about using accumulated knowledge about the person driving.
I can’t help but feel my descriptions getting more and more similar to just describing a competent person. And, I’m aware I’m being idealistic and what I’m OK with won’t be a product any time soon.
I guess it would be fully on device, encrypted at rest and have a perfectly good memory of our conversations and it would be tactful in bringing in knowledge into conversations. I dunno, I’m just describing the ideal personal assistant AI. And many people would make it a companion. And… yeah, anyway. Surveillance capitalism and pervasive advertising is bad.