65a

12 ms·

91.
by 65a 3y ago
Hasn't VAG had steer-by-wire for nearly a decade now?
92.
by 65a 3y ago
I reported maybe the first instance of that (#15275), and lost my home pool (but was able to get my files back). It was not a good weekend.
93.
by 65a 3y ago
Note that dungeon (the source port from the mainframe game that eventually became Zork) has been open sourced for much longer. At some point it was machine translated from F77 Fortran to C~90, and comprises some of the most terrifying code
94.
by 65a 3y ago
I agree I probably should have said strongly typed instead of safe, as yes, if you dereference a pointer to nil you are going to crash. That being said, I do think "possesses an untyped nil" is a pretty far cry from "not type
95.
by 65a 3y ago
It is a type safe language, not exactly sure what you're hinting at here.
96.
by 65a 3y ago
Go generaly doesn't used shared memory and concurrency, or at least it's been considered an anti-pattern: https://go.dev/blog/codelab-share
97.
by 65a 3y ago
I have here in my hand a long list of real world outcomes of these natural intelligences, and I assure you the risks cannot be understated . Never before has such cruelty, evil, or destruction been wrecked upon the innocent, and I assure y
98.
by 65a 3y ago
> The basic idea is that AI is the opposite of software; if you publish a model with scary capabilities you can’t undo that action. I find this a bit naive. Software can have scary capabilities, and has. It can't be undone either, b
99.
by 65a 3y ago
Probably officially dropped, but gfx900-series works fine on ROCm 5.7 here, which actually goes back to vega10 (IIRC mi50/mi60 are vega20).
100.
by 65a 3y ago
I grabbed a 32gb mi60 on a lark, because at the time it was ~$300. It works fine with recent ROCm etc, but it has the following potentially severe drawbacks: 1. It is not a recent GPU, so don't expect blazing inference speeds 2. It is
101.
by 65a 3y ago
My 2016 desktop had 22 cores and 44 threads. You can have the same processor for < $200 on ebay right now.
102.
by 65a 3y ago
UDBOM DIFTY SXWLB EQAUJ BUJYX JVRHR EGVVM VMFNZ AORRI WGLPW YJNA OMOZO DAPLE VGZNA TEBKM XCPEA AQSZV OYGYQ RWWCX Z
103.
by 65a 3y ago
You would want '</s>' as a stop token.
104.
by 65a 3y ago
You can also use high-Q antennae, if you have a frequency of interest. For instance, on HF (shortwave) a tunable small loop antenna does a whole lot better than a wire antenna in a high-random noise environment. You may still end up chasing
105.
by 65a 3y ago
You were swayed by a hypothetical and a 19th century doctrine? I'm sorry to hear that, but it doesn't absolve Russia of responsibility for invading a sovereign country.
106.
by 65a 3y ago
I can't hear you. I'm using the scrambler.
107.
by 65a 3y ago
I'm pretty sure Russia started that by invading in 2014 and then 2022. Ending wars by giving invaders other sovereign countries doesn't work so well...it certainly didn't serve peace last time.
108.
by 65a 3y ago
These work completely fine on linux.
109.
by 65a 3y ago
Confabulation seems better aligned to neuropsychology, as far as I can tell: https://en.wikipedia.org/wiki/Confabulation
110.
by 65a 3y ago
It's a feature not a bug. Maximizing banality is being cast as safety to avoid discussing the harder problems long term. I agree that any new technology comes with risks, including serious ones, and humans should get better at managing
111.
by 65a 3y ago
If your argument is that users of mathematics or computers are responsible for their actions, I agree with you. My comment is about the researchers arguing (in effect) that no one should have a computer because they might do Therac-25 , wh
112.
by 65a 3y ago
I haven't seen enough papers about safety for computers or mathematics in general. Has there been any progress on preventing them being used for anything harmful? Could we possibly only allow an elite few to use them? (For the sake of
113.
by 65a 3y ago
Not quite, there are 128GiB AMD instinct parts, though the prices are very...enterprise: https://www.amd.com/en/products/server-accelerators/instinct...
114.
by 65a 3y ago
I was running llama on a w7900 a month ago, with 48gb of VRAM and excellent performance. ROCm support got a lot better really recently.
115.
by 65a 3y ago
ROCm works fine on my 2016 Vega Frontier edition, for what it's worth.
116.
by 65a 3y ago
As an owner of some Sapphire Rapids parts, let me just direct you to: https://edc.intel.com/content/www/us/en/design/products-and-...
117.
by 65a 3y ago
Agreed...the main gap is support on consumer and workstation cards, which is where nVidia made headway, but that is starting erode super recently. ROCm works pretty well for me, I have had a lot more problems with specific packagers than th
118.
by 65a 3y ago
I literally gave away my last laptop with a discrete nVidia card because it wasted so much of my time.
119.
by 65a 3y ago
Will they? It's pretty clear that the trend is toward smaller but more efficient models, which further reduces their moat.
120.
by 65a 3y ago
There are no community fine tuned models here either, many of which outperform the open source base/chat models on many tasks.

More ›