🧵 Untitled Thread
Anonymous at Mon, 8 Jul 2024 19:53:00 UTC No. 16273150
>when we were told that Blackwell's ambitions were beyond the limits of Physics, the engineers say "so what?"
https://youtu.be/odEnRBszBVI?t=281
How the fuck does >>>/sci/ respond to that without sounding mad?
Anonymous at Mon, 8 Jul 2024 19:53:46 UTC No. 16273152
>>16273150
>Blackwell
who?
Anonymous at Mon, 8 Jul 2024 20:14:13 UTC No. 16273167
>>16273150
How much for one board (cpu included) ?
Anonymous at Mon, 8 Jul 2024 20:16:26 UTC No. 16273173
>>16273152
I would assume it refers to David Blackwell, pretty much the only """black""" mathematician who's done anything significant.
Anonymous at Mon, 8 Jul 2024 20:38:05 UTC No. 16273196
>>16273150
fine, but will it run Crysis?
Anonymous at Mon, 8 Jul 2024 20:45:51 UTC No. 16273211
>>16273196 It will cause a world crysis for sure.
Anonymous at Mon, 8 Jul 2024 23:21:14 UTC No. 16273394
>>16273167
$80,000
Anonymous at Mon, 8 Jul 2024 23:39:46 UTC No. 16273416
>>16273394
>80k
The whats his name can go fuck himself
Anonymous at Tue, 9 Jul 2024 00:17:49 UTC No. 16273458
>>16273416
It is not for your computer games manchild. It is for professional work where the customers are big companies
Anonymous at Tue, 9 Jul 2024 11:28:51 UTC No. 16274017
>>16273458
>professional work
>using generative 'AI'
>the primary use case is spamming slop, with little evidence for real applications beyond it (no, the company slapping "AI-powered" on a product doesn't count - using regular ol' logic switches is STILL more efficient for everything EXCEPT slop spamming)
>already hitting hard scaling limits that hardware can't solve (training data availability and quality, for example); best the hardware can do is output slightly faster
>we're stuck in the dystopia where midwit tech CEOs took empirical proof that the Turing test wasn't worth a damn and instead went whole-hog into declaring that a perceptron is intelligent as soon as the threshold was crossed
>a perceptron is basically a primitive motor cortex at best - doesn't even have feedback mechanisms, which real motor cortices do
>we'll be stuck in this local maximum for a few years until the bottom falls out and AI research gets slashed, recovering a decade or so later for a new hype cycle... again.
>because this has happened before
>WITH PERCEPTRONS
fucking hell, we're never going to get AGI, are we?
Anonymous at Tue, 9 Jul 2024 11:38:35 UTC No. 16274023
>>16274017
Yeah I'd rather have 4k videos of Hitler painting over the datamining and forced retardation of the internet