Earlier today I received an email from Proton with an annual survey. Among standard questions there was a significant amount of AI related questions, e.g. mail assistant. Does it mean Proton is looking into AI?
Not knowing what the questions were, I would guess they want to know if users are actually interested in it. Presumably if everyone says they want one they’ll invest, but if everyone says they don’t then they won’t.
@smeg@fart_pickle Hi everyone, and thank for the feedback! We send out an annual survey to align with the community.
Proton currently uses AI in Proton Sentinel, to help protect your account from takeovers, even when an attacker has your password. This is optional for subscribers and needs to be manually enabled.
Any future applications will depend on community interest and demand.
I would be happy to see privacy respecting AI assistant in Proton products, especially Proton Mail. English is not my native language and every now and then I need to use 3rd party services to compose a professional sounding email. Having this built into Proton Mail would have made my life easier. Also, having a feature that detects and labels faces and features in the pictures stored in a Proton Drive would be an amazing addition.
@fart_pickle@protonprivacy
If Proton develops a "feature that detects and labels faces and features in the pictures stored in a Proton Drive," I will IMMEDIATELY extract every photo and file of mine from Proton Drive, and then CLOSE my Proton account.
Having that feature would probably require users to provide their password to the system performing the 'facial recognition' scan. That would be a HUGE vulnerability.
If I ever want my privacy destroyed, I'll use Apple, Google, Facebook, etc.
This echoes my thoughts. I really suck at writing emails in my native tongue, and having something I can prompt to write an email for me would be useful. I don’t want it to be trained on my emails to replicate my style or whatever, I’d just want an LLM on the side that I can prompt to write for me. I’ve done that with other LLMs in the past without having to integrate it anywhere.
I don’t particularly care for facial recognition. I don’t take photos of people.
There was a question about AI and privacy. Don’t remember exactly what it was but it was something like “What do you care the most about integrating AI into products” and most of the answers were about the privacy.
There are foss ai models that can work offline, so it’s not completely out of the question to develop something that respects privacy. But IMO not entirely sure of that’s the best use of resources, unless this doesn’t slow down development of other improvements
Ollama seems pretty good. I’ve been playing with it recently and when combined with openwebui, it makes pretty nice local chatgpt. But I agree, there are more pressing issues I would like to see fixed/implemented.
Add comment