Chrome Is Downloading A 4GB AI Model Without Your Permission

Argent Stonecutter

Emergency Mustelid Hologram
Joined
Sep 20, 2018
Messages
7,504
Location
Coonspiracy Central, Noonkkot
SL Rez
2005
Joined SLU
Sep 2009
SLU Posts
20780
Would actually hope that any application running a local language model is doing the same thing.
 

Noodles

The sequel will probably be better.
Joined
Sep 20, 2018
Messages
6,061
Location
Illinois
SL Rez
2006
Joined SLU
04-28-2010
SLU Posts
6947
Would actually hope that any application running a local language model is doing the same thing.
Yeah, I agree. I do wonder if this is only doingnthis on machines with an NPU on it. Because if you can't really run AI locally, what would be the point?

I have seen this story goingnaround but it kind of reeks of someone not realizing they asked for this to happen and making a huge deal out of it.

And I hate Google, I don't trust Google.
 

Argent Stonecutter

Emergency Mustelid Hologram
Joined
Sep 20, 2018
Messages
7,504
Location
Coonspiracy Central, Noonkkot
SL Rez
2005
Joined SLU
Sep 2009
SLU Posts
20780
Well, I wouldn't characterize it as something they asked to happen, unless you count "Chrome started adding chatbot features and they didn't figure out how to disable it".
 

Noodles

The sequel will probably be better.
Joined
Sep 20, 2018
Messages
6,061
Location
Illinois
SL Rez
2006
Joined SLU
04-28-2010
SLU Posts
6947
Well, I wouldn't characterize it as something they asked to happen, unless you count "Chrome started adding chatbot features and they didn't figure out how to disable it".
Probably not, but there have been similar things like this where it was like "Oh, it was a checkbox that everyone blew past on this conformation screen".

A recent similar one.

There was all this "Microsoft is storing password in plaintext" or something. But there were a lot of security people basically pointing out, everything does this, it has to do it this way for it to work, ita not permenant, and most importantly, in order for this "exploit" to be done, someone would have to have in person access to a compromised machine, so they would have way more useful access to more dsngerous stuff anyway.

Something along those lines. A big todo about something that was essentially, BAU.

Maybe this Google thing is a big deal, but it also kins of feels like a big "AI Bad" moment that may be nothing. Like we find out later it only happens on AI Chromebooks or something, specifically designed and advertised to need and download a local AI model.
 
Last edited: