This reminds me of another tech bro many years ago who also thought that expertise is overrated, and things really aren’t that hard, you know?

lmao, what’s his lesswrong username?

it’s worse than that, you should probably take number of adults (as in, 18+) as base here, and it’s 78% of them (267M), according to first random source i’ve found, so it’s closer to 34%

from what i understand, american anomaly is that they take debt like that even when not strictly necessary in order to pump up their credit scores which might be useful later, but even then, 9% of population relies on going to loan shark the app in order to get food, absolutely nothing to look at here, move along,

These sound like some random plankton orgs, you can just as well pick a name and start your own

the thing is in specific ways much older, i think that if you want to stretch it the first people you can credibly accuse of being ai bros were alchemists trying to cook homunculus, or trying to get infinite knowledge out if philosophers’ stone

For now

this is how 238Pu ceramic pellets for space probe generators look like, no fission required just alpha decay. If it was fission, it wouldn’t need to glow like this entire time because you can just turn it off

i don’t know, maybe softbank counted on openai not being able to go for-profit (which is not IPO, so it’s not like they got more transparent or anything). that deal was signed long time ago in bubble terms, and it went like this (or so i heard): first tranche was 10B from softbank + up ro 10B from other investors, they couldn’t get all that money from other investors and that’s why it was 18.5B and not 20B. that was some half year ago? even back then softbank didn’t had the money and had to sell part of arm (iirc). now they had to supply the remaining 22.5B, idk if it’s all softbank, or whether they would get out of that deal if they could, or whether they will continue pouring money there (we can assume all these dcs are vaporware) can openai find other suckers? possibly, but scaring away a reliable one would be bad for business

new zitron: ed picks up calculator and goes through docs from microsoft and some others, and concludes that openai has less revenue than thought previously (probably?, ms or openai didn’t comment), spends more on inference than thought previously, openai revenue inferred from microsoft share is consistently well under inference costs https://www.wheresyoured.at/oai_docs/

Before publishing, I discussed the data with a Financial Times reporter. Microsoft and OpenAI both declined to comment to the FT.

If you ever want to share something with me in confidence, my signal is ezitron.76, and I’d love to hear from you.

also on ft (alphaville) https://www.ft.com/content/fce77ba4-6231-4920-9e99-693a6c38e7d5

ed notes that there might be other revenue, but that’s only inference with azure, and then there are training costs wherever it is filed under, debts, commitments, salaries, marketing, and so on and so on

e: fast news day today eh?

if you think it’s stupid, it’s not as stupid as it sounds, it’s worse. they also sold some tmobile stock and took debt backed by ownership of arm. it’s like they instinctively get rid of pieces of ai bubble that retains some money and hold to pieces that are black holes

things that bounce from ionosphere. shortwave radio, AM broadcast, some military communications. couple of radiooperators will take a break for a day or two until space weather gets calm. unlikely that satellites would get fried

if you’re lucky, you might see aurora tomorrow, even if you’re in area that usually doesn’t have them (easier for southern hemisphere)

my read of the situation is that it’s another phenomenal softbank L even if they timed this sale at the top of the nvdia valuation. if they thought it’s a bubble popping soon, they would try to get out of openai deals, but they’re doing the opposite. most immediately, they need money to dump 20B-ish into openai by end of the year, triggered by that no-profit transition, and it’s money that they apparently don’t have. that their stock dumped like 15% in a week probably didn’t help either

These people aren’t looking for scientists, they’re looking for alchemists

can you pull ethernet cable along power cables, wherever they are?

regular ethernet should work on this kind of distance, but it means digging

and only feels cool and useful when you’re learning to write

Mate do i have just the right thing for you, but it requires some soldering. It’s also probably cheapest solution working over longer range than you need

First you need two directional antennas. Use this https://lea.hamradio.si/~s53mv/wumca/cup.html the 13cm design specifically. Design of the dipole element is on another page https://lea.hamradio.si/~s53mv/wumca/sbfa.html They’re using hard to get semirigid coax but you can really just use common RG178 with braid tinned to make it stiff. This way you don’t have to leave D section they way they did, you can just solder core to the shield at the end while preserving total length (or ~1-2 mm less, because wifi is slightly higher frequency; 53-52 mm total). That dummy cable thing can be just any stiff piece of wire. Good way to get this would be getting a pack of u.fl-SMA pigtails, which you can also use for connection.

You also don’t need special aluminum housing like they do, cookie tin of the right size would be sufficient, or any other container of similar nature. If you can’t weatherproof it, putting it inside on windowsill is also fine

Then, plug TL-WN722N into it, or some other single-antenna thing, and you’re set. This one connects over USB and has removable RPSMA antenna, so you can connect it easily with correct cable (SMA plug - RPSMA plug)

to your new directional antenna. This thing works well over 200m distance, provided clear line of sight, and probably more than that

here if you need anything over certain power (6kW; depends on country i guess) you need a three phase installation, and even if you get single phase, it’s really handled as three phase split between single phase customers (a block gets three phase supply, then splits flats in three groups, each group gets connected to one phase). this gets supplied by a distribution transformer that might serve somewhere around 200 people per (in residential areas)

i understand that sometimes americans also get distribution like this, with 208/120 three phase coming from substation, without 240v available

american version would probably only have two phases at best, and possibly just one

wait, how do you route cables in there? is there just a massive bundle right through the middle?