When I read about Chrome shipping local weights in every computer I can’t help but be reminded of the Great Cannon, where a MITM code injection by China caused end-user browsers to launch a DDOS on GitHub back in 2015. What could an AI adversary do with a fraction of Chrome’s local models, I wonder.
Why do we think that?
I'm not getting specific because I think people are ragging on Chrome too much or whatever.
I'm really alarmed that Chrome could eat 100s of GBs a month and people are handwaving about the cause(s?) in a way that shades rather than illuminates.
What else, going on with this guy's computer, would cause that volume of traffic to immediately stop as soon as he "disabled everything on that settings page"?
It may not be happening to you if you've got all that crap enabled, but that doesn't mean it wasn't happening to him. Companies love to single out subsets of their userbase and secretly experiment on them. When something goes wrong users get gaslighted by other people who say "I never see that happen!" or "It works fine for me!". Who the hell knows what Google and their AI was doing on this guy's computer. I'd put more trust in a random web developer on the internet than I'd trust Google.
My hypothesis is this guy installed so many shady ad blockers and VPNs that were using his rig as a proxy, and the cessation of such traffic, whether permanent or temporary, was purely coincidental.
But also low enough that most of us wouldn't actually notice
The problem here is without a deep dive you just don't know. Google likes A/Bing the shit out of their software. Google loves setting all kinds of different options that may have to be replicated exactly to get the same behavior.
This isn't like support back in the old days where you just look at your software and find the bug, you have to replicate the environment to see what's up.
Without knowing, you don't accuse.
Sometimes I notice my PC being very very slow... The Chrome Task Manager (SHIFT+ESPACE) shows that "On device" IA is taking 300% of the CPU (laptop with only low-perf Intel integrated GPU). I can barely type text in an input box: 1 second delay after each keypress.
Disabling the option "On device AI" in the same way than on OP's screenshot doesn't solve this issue: this "On device" process still comes back again and again and I have to kill it manually.
Did someone experience the same?
(Linux)
Chrome: `/etc/opt/chrome/policies/managed/disable_ai.json`
Chromium: `/etc/chromium/policies/managed/disable_ai.json`
```
{
"GenAiDefaultSettings": 2,
"AIModeSettings": 1,
"GeminiSettings": 1,
"GenAILocalFoundationalModelSettings": 1
}```
Fast used to be a feature.
I thinking this is just a wave of hot-patch-slop that will pull back soon-ish, if the feedback loop works. Many times the folk making these decisions are so far removed from customer complaints, the there is any complaint (because many users don't bother)
chrome used to be the simplest browser, with a minimal interface that was lean and fast...
I bounced off chrome very quickly, tried it again and again over time, never stuck with it.
I’m no fan of a Google monopoly but I still prefer where we are today to those days.
I have resorted to running Chromium running for Teams, but that's all I use it for. It's the only thing that doesn't work for me in Firefox.
You can disable Chrome AI features forever, by switching to Firefox!
Point to the specific AI features which are wasting bandwidth and quantify how much.
It doesn't have the AI features, which improves the product quite a bit.