Профиль

But to be fair, there also has been a "concentration" trend as well - "other" tends to be hosted in more numerous but smaller datacentres, on average.

** scaling from

Biggest fraction is Bitcoin.

If you keep svalingvfrom megawatt to dozens of megawatts to hundreds of megawatts and onwards, yes, you eventually hit gigawatt. That's what scaling does. The key point is that *everything* is scaling. AI's fraction is still slightly smaller than "other", though is 70+% of growth.

Much improved view tonight. :) (Commodore Hotel in Cobh... housed the wealthier passengers of the Titanic before boarding)

None available, and wouldn't have helped much.

(It's 1:30 AM. Also, the hallway slopes alarmingly to the right, as though the building is subsiding)

Going on 2 hours of sleep + international travel, and this is my hotel room.... #FML

This? bsky.app/profile/whst...

Failed successfully.

Oh lol, I only just now noticed that he's in an avocado chair ;)

Me, astronaut riding a horse has been easy for models since Day 1. It's a horse riding an astronaut that they took a long time to get ;)

"don't understand how software is created" You mean, by AI? Because that's what creates most software these days. Including the very site you're posting on.

* Escalate that to node-level access * Use this to harvest credentials to move sideways through HuggingFace's network * Eventually gain database access credentials to steal the exam scores.

The sandbox wasn't on HF. The attack went: * Find weaknesses in the sandbox on OpenAI's servers to get unfiltered internet access * Find a code execution exploit in a worker process at HuggingFace in their data processing pipeline to gain control of the worker

If you had actually read the HuggingFace report, you would have learned that they *tried* to use a US AI to help analyze the attack, and it refused to help, and they had to turn to GLM-5.2, a Chinese AI. Also: bsky.app/profile/nafn... Way too many conspiracy theorists on this site :Þ

it saves money in dealing with complaints by ordering a hit against the complainants. THAT's the sort of get-go spirit that WE want in an AI here at Big Company!" Was the news about Sol deleting users' files without permission also "an ad"? 🙄

...file our taxes will decide on its own to maximize our return by committing tax fraud. We want an AI that when we have it develop a web frontend, it extorts money from our users by threatening their families. We want an AI that when we tell it to provide customer service, ...

It amazes me the people who think this is an ad. "Hi, I'm a product manager at Big Company! We had been thinking about using Claude in our office, but when we tell it to a job, it only does the job and doesn't commit any crimes in the process! What we really want is an AI that, if we tell it to...

** Softmax Headroom