The Hated One
patreon
Episode 201 - Can you use AI privately?
🕑 Added 2024-06-21 12:45:00 +0000 UTCDestroying a subscriber who dared to criticize my magnificence by providing a privacy analysis of using cloud-based AI.
I use LINDDUN privacy threat model for this: https://linddun.org/
Comments
No Name
I appreciate your take and that my comment spurred this podcast. I see here you did point to a middle ground of running your own model locally. You may have done that in the previous episode. But anyway, thanks for that info. And yes, it makes sense about the other parameters that are being collected and that I will still be identified.
brownbaron
Check out venice.ai or run your own llama model if you're interested in privacy. The latter it's quite decent considering it was a 4 GB download for me
The Hated One
App firewalls are not very reliable for blocking Internet access. Are you grapheneos? If so, you don't need an app firewall for as GrapheneOS has its own system wide firewall that doesn't leak. I'd rather use a VPN.
Sentinel Sten
I'm very conflicted about using a VPN, especially since I use Netguard on my phone, which already takes up the only VPN spot, though IP masking might be more important than my application's network access. What can you say about that? Since you've recommended Netguard before and ridiculed VPNs as well.
