shdjhdfh

Right, he should have phrased this differently, but everyone is ignoring that he explicitly says "from a global allocation perspective" immediately before it, making it clear the "terrible" part is from a very narrow perspective, not as s general rule.

He doesn't say anything about individual users. It's left unclear exactly what he means when he talks about low utilization inference using DRAM. I figured he was talking about companies building clusters. I see way more people using old GPUs for local inference than buying 15-20k Mac Studios, or RTX Sparks. I've seen plenty of stories about how companies are building clusters because they want to run open source models, partly for privacy purposes, partly because they're afraid of a rug pull, partly because they want to control when they move to new models, etc, etc.

All of this, plus it knows what agent is running in which windows and what session that agent is working on. If you have to reboot, it can reconstruct all your terminal setup and restart the agents on the right session automatically.

It'd be trivial to prove and yet it has never been proven. I'll give you three guesses why, and the first two don't count.

I doubt any TV company will ever do this, for two reasons. The first is simply that it would be obvious due to FCC filings. The second is that 99% of their customers hook their TV up to the internet so they can use streaming apps on their TV. They're very unlikely to eat the cost of putting a cellular model in every TV, take the PR hit for doing so, simply to capture the ACR data of the 1% of their customer base that is most likely to figure out another way to disable it.

Whereas your CPAP machine is likely a high margin device and the reason it has that modem is because insurance companies require compliance data in order to continue providing coverage for replacement supplies. Resmed or whoever is happy to bundle in the cost of the modem + service in order to keep selling high margin supplies.

There really wasn't a lot of new information in the Gamer's Nexus video. Most of what people think is new is people conflating what is possible vs what LG does and people conflating what is possible on a rooted TV vs what LG does. The whole video is very long and because it's so poorly organized and deceptively presented, it looks like LG is worse than everyone else. Rally they're just doing what we've known for a while that all TV manufacturers are doing.

We don't really have any idea what was in his sessions or why he was suspended. It might have been a false positive in their automated system, or maybe he ranted at the model about how he was going to "destroy" Samsung with his findings and the automated systems took that literally. There are tons of examples around of people using models for reverse engineering, so we know it's unlikely that alone was the reason.

If they wanted to know if you had dogs or babies and wanted to spy on you, they'd just transcribe the audio. People that have dogs or babies talk about them. Going to the effort to detect these and advertise it as a feature so you can secretly use it for marketing doesn't make a lot of sense.

You should assume it will not work unless you test it regularly. That's a big part of why having active/active multi-region is attractive, even though it's much more complex.

My guess is that skywhopper is correct. We're only able to see the tail end of the attack, but the repo was likely compromised in some way.

Another thing to note, the AI angle on this is nonsensical. The commit could have just as easily done many other negative things to the system without AI as a layer of indirection.

Your article breathlessly blames AWS for being reckless while having no real facts about the compromise. The whole thing reads like click bait.