his perspective
“…have to reduce spending to live within our means.” – Richest guy on the planet
his perspective
“…have to reduce spending to live within our means.” – Richest guy on the planet
Fish and wine for everybody!*
https://x.com/america/status/1850346392172048635
*After you’ve suffered a little. Maybe after you pay my way to Mars. And when I think we can afford it.
https://x.com/MarshallCohen/status/1851740660052955505
Won’t be too inconvenient. I think he’s doing the hardcore there in PA.
This is a scary thing to hear because we don’t really do “temporary hardships” in exchange for “long-term prosperity” in this country. All that matters is the here and now. And if you don’t deliver right away, you get voted out of office. Unless of course, you’re a dictator, and you just declare yourself the winner even when it’s obvious that you lost.
The temporary hardships for long term prosperity is pretty much the degrowth agenda, which democratic centrists have been pushing as politically difficult, because it is. It’s also what Great Britain did during the Tory tenure and resulted in 10+ years of no GDP growth.
Won’t show. Will be zero consequences for it.
I bet there’s gonna be some sternly worded letters though!
There’s something weird/interesting going on with xAI’s Colossus cluster.
I’ve read in a couple places that the cluster is wired together using Ethernet rather than InfiniBand, which is the universal standard for AI training clusters. I don’t really understand it, but my sense is that InfiniBand is substantially more complex to install at scale because it’s designed to provide more direct connections between nodes, which requires a flatter and/or more densely connected network. The advantage is that it offers lower and less variable latency. This is important for AI training because there’s lots of back-and-forth communication needed to decide each update of the model weights during training.
This decision is possibly related to recent comments from Jensen Huang about how incredibly fast the Colossus cluster was built, and that xAI was able to install 100k GPUs in 19 days, a process which would normally take years of planning and a year of build time at other places. Pure speculation, but were they able to build Colossus so fast because they took a shortcut with Ethernet?
This sounds very arcane, but it’s important to realize that 100k H200 GPUs probably cost something like $2-3 billion, and building an entire datacenter around them is a major capital investment too. I see three possibilities:
Based solely on the fact that it would be the most interesting answer, I’m inclined to think that it’s (3). I’m very very interested to see what happens with this though.
Three does seem like the most likely situation.
Musk’s attorneys filed documents Wednesday night in the U.S. District Court in Philadelphia, a federal court, requesting the matter be moved to their jurisdiction – and out of the local Court of Common Pleas. A judge granted that transfer.
Attorneys for Krasner said they would contest that recommendation.
“We will proceed to federal court and we will address the issues there, and seek to have the matter remanded back to the state court. After all, this is a case that involves state law issues and I’ll leave it at that,” said attorney John Summers.
Ah! Well. Nevertheless.
I’m stunned.
Consequences!
Normal people get slapped with all kinds of fines and even jail time if they blow off court appearances.
Idk how they would get this information. I don’t recall being asked for an employer name when I contributed. Maybe I said NA and forgot. Anyway, this is pocket change compared to Elon.
Magnificent film.
Yeah. I kind of want to find it streaming somewhere now. Didn’t expect Marion Ross looking hot in the trailer.
I have it on dvd, pull it out at least once a year. Final scene is just perfect.