From the SCMP article you might get the impression that the various figures all refer to the same GPU cluster, but in the paper itself it's very clear that this is not the case, i.e. the 213 GPUs in the smaller cluster are not serving 1.35% of the requests in the larger cluster. Then if you want to scale it, you have a choice of different numbers you could scale, and each would get different results. Since they're constrained by the limited number of different models a single GPU can serve, I think scaling by the number of models is the most realistic option.
Not how soon will the seas rise, but in which direction are they trending? The world is not static - it's dynamic, always changing. Long before human civilization, the environment was getting warmer, then cooler, then warmer, etc.
Instead of freaking out about it, work to understand how we as a civilization can make it through the changes. Attempting to hold the environment stable is like plugging dikes with your finger.
Humans are causing a change to the environment drastically faster than any normal ebb and flow. In the same way that you can get from 0-60 either by pushing on the gas pedal, or being hit by a train, rates of change matter a lot.
I don’t think anyone here is advocating for freaking out about it, and I do not see any suggestion we should that you’re replying to.
Also, it’s pretty clear from the IPCC report that holding the environment stable isn’t an option. At this point about 0.5m minimum sea level rise this century is probably locked in. However a 1m rise might not be.
So yes, the issue is that either way policy planning needs to be considered in the long term.
Both sides do this: the last administration had tech giants censor anyone that that promoted the lab leak theory, or any alternative to treatment other that the new vaccines. Parents interested in managing their child's exposure to sexually related material were put on watch lists.
It's not one side or the other - any group with authority has to be watched closely and rebuked when they try to expand their power.
I've heard that too. But, from my experience, it's mostly insecticides: people want mosquito-free and flea-free (and ants, roaches etc) yards, and spray/spread poison that kills them. Unfortunately, it kills everything. I cancelled my yard service and noticed that birds started coming back, chasing the food supply. Butterflies cover my lantana, and I see fireflies at night.
I live in Minnesota and the struggle is tough. Either my yard is completely unenjoyable for the family due to mosquitoes and deer flies some years or I kill 50 insects that I don’t want for every one that I do.
I put up deer fly traps every year and that helps, but this year in particular has been awful for both them and mosquitoes. Luckily for the bugs my sprayer is broken and I haven’t had time to fix it.
I deliberately avoid the place on our property where fireflies are, on the years that I do spray. We really need a better solution.
Just to be clear, if we go outside right now we instantly have multiple deer flies trying to land on us and bite, which is only a distraction for the mosquitoes. Without swatting I’d probably get a mosquito bite a minute midday or maybe 5-10 a minute at 7pm.
You can target mosquitoes more directly with mosquito dunks. It's a block of dehydrated bacteria that produces a toxin believed to only impact the larva of mosquitoes, fungus gnats, and black flies. Basically you make sure all standing water in the area is drained or has a mosquito dunk in it. You can make buckets full of yard debris and water and a mosquito dunk to attract mosquitoes to lay their eggs that then get killed.
You can also try to increase the number of predators. Create habitats for bats, dragonflies, mosquito larva eating fish, etc.
I've found picaridin lotion to be much less annoying than DEET spray but still effective this year.
Yeah, I’ve used mosquito dunks experimentally but we have a pond behind our house and in front of it, and a lake nearby. It’s Minnesota, after all. I did try to chuck some dunks into the ponds but they have a lot of cattails in the way so I was unsure if they even made it in the water.
Our attic was such a good habitat for bats we had to do something about it ;). But we’re also surrounded by woods.
It would be great to help keep CCT skills fresh and current - rotate them around with 'deployments' to civilian facilities. It'd be a win all the way around.
It takes 1-3 years of facility based training in order to qualify as an ATC for a specific tower. A military enlistment is only 4 years. CCTs would end up doing their entire enlistment at one facility leaving no time to train as a CCT. You'd have to hire another person who's job is 100% CCT, at which point, why not just hire an ATC?
It sounds like a nice idea, but the only common factor between ATC and CCT is the certification and some fundamental core training -- everything else is super nuanced specific to each scenario. Some CCT who's specialized in deconflicting a stack overhead in wartime can't just waltz into ORD tower and say "all right boys and girls, go ahead and take a break, I've got this." Each requires domain-specific experience.
And to add on what others have said: yes CCTs represent a pool of proven ATC candidates, but depleting that pool just to knee-jerk a short term-ish solution creates an equal problem for the military -- and it's a hell of a lot harder to recruit adequate candidates for CCT. For example, they have to do like, lots of pushups...
"A paper presented at SOSP 2025 details how token-level scheduling helped one GPU serve multiple LLMs, reducing demand from 1,192 to 213 H20s."
Which, if you scale it, matches the GPs statement.