pull down to refresh

The excitement around Google’s Gemini AI making its way to Android is almost exclusively based on its sexier abilities, but it’s the boring stuff that has really piqued my interest.
Or, at least, the potential for the AI to do boring stuff. I asked Gemini to find emails from British Airways in Gmail. Not only did it find the messages, but it categorized them (T&Cs updates, upcoming trips, account and privacy) and summarized relevant emails so I could quickly find the information I needed. To me, this was more impressive than anything else I have seen Google demonstrate.
Question really is, will it be possible to use such assistants without any privacy issues...? The reason why I personally do not use ANY google or samsung or whatever assistants is, because I don't want any company to analyze my data/behaviour.
reply
Of course not - this is the trade off you have to make in order to get use of a functionality which needs data inputs and then analysing them.
How can you expect it to assist you without collecting and analyzing the data it needs to do its business?
It's another topic of whether it should persist it after the assistance is finished...
reply
The software needs data inputs, the company doesn't.
So there are 2 solutions for this:
  1. Run it locally. But there is an issue with computing power (it get's better, but I don't know what the limits will be).
  2. Send data (encrypted of course) to a stronger machine (server), let it do the analyzing and output creation, send results back (encrypted of course) and then delete everything on the server. But 2 issues here:
  • You have to trust that data and results will really be deleted.
  • You have to trust no data is leaked anywhere between.
reply
If the data is strongly encrypted (only you can decrypt it) why do you care if it's deleted or stays on the server as it is practically unusable garbage for them?
reply
The software on the server needs to be able to decrypt the data which users are sending, otherwise it cannot analyze.
reply