"coprocessors, but matrix-math specific"
the various *PUs are "things that help a lot of ML models run faster" sidecar chipset designs
it's actually kinda hard to get concrete details, afaict. I've had a bit of a look around for silicon teardowns and shit, and haven't really found any good ones yet
ooh wow, the speed of progress. such a well known indicator of something definitely being good, and never at all being abused by ghouls!
sometimes I wonder if any of you dumb fuckwits ever read any historical perspectives at all
"so true bestie" says the dipshit, brushing under the carpet all the notes about overt efforts and capital expended by monopolist captive operators
what in the utter fuck is this post
.....I was unprepared for reading this post
remarkably, your post is one of those "tech takes" that we hang around the watercooler for
although I'm not sure you'll get why
broadcom continuing to make vmware even more unappealing
bit of a fucking weird thing to do, too. guess they’re going full oracle?
wtf is going on with that touchpad - is it a tap calculator input?
longer: dragon dictate and similar go back to the mid 90s (and I bet the research goes back slightly earlier, not gonna check now)
similar for TTS
nah, just call a fuckwit a fuckwit. even jokingly giving them breathing room is something they know how to abuse.
the fuck? please go be weird somewhere else