krasin

22 ms·

211.
by krasin 3y ago
> Am I reading that correctly? For $50 you get ~1,000 transistors worth of die space, and it comes in a package? That's correct, although it's $100: $50 for the space on the group-run chip and $50 for the physical stuff: the ch
212.
by krasin 3y ago
The easiest way to try Skywater PDK is through TinyTapeout: https://tinytapeout.com/ -- highly recommended.
213.
by krasin 3y ago
Interesting! Can you please name some of these old games made with Gaussian Splatting? I would be interested to play, to get a sense why polygons won in that round (and likely to lose in this one).
214.
by krasin 3y ago
Nice, Apache 2.0 license. Thank you, Replit! https://huggingface.co/replit/replit-code-v1_5-3b
215.
by krasin 3y ago
1) It should be possible to build a mostly passive dry cooling tower. It will cost more to build, but the cloud companies are not particularly struggling. 2) Electricity is much more portable than water. Droughts are regional and there&#x27
216.
by krasin 3y ago
If the end result is a lot of freshwater lost, it's still very bad.
217.
by krasin 3y ago
I agree. Which is why full water recirculation for datacenter cooling should probably be a law.
218.
by krasin 3y ago
> The water would need to be cooled before reuse, which could look like a car radiator and fan. Dry cooling tower is the word, I believe: https://en.wikipedia.org/wiki/Cooling_tower
219.
by krasin 3y ago
Thanks. That's a little better, because it has a higher chance of turning back to rain somewhere around (or at least making air a little more moist).
220.
by krasin 3y ago
In the question of datacenter water usage, its application (AI or not) is actually not very relevant to the problem at hand. I challenge the fact why (if true) the datacenters even need "new" water, instead of having a closed syst
221.
by krasin 3y ago
Is it really true that datacenters use freshwater once and then just discharge it to sewer? If so, the criticism of cloud companies (including but not limited to Microsoft) is valid. Especially that it's reasonably straightforward to b
222.
by krasin 3y ago
The immediate flashback from reading your message: J3QQ4 (but that was probably Windows 98)
223.
by krasin 3y ago
An alternative is to write most of the program in Python + JAX + implement a few custom XLA ops in CUDA / Triton. That way, the program is very readable and can interoperate with the larger ecosystem, while still being fast to run.
224.
by krasin 3y ago
> Regardless, s-2 glass fiber is probably a better choice I would appreciate if you can elaborate on that. Not being a material scientist, it's not obvious to me which option is a better choice here, but I am curious to learn.
225.
by krasin 3y ago
Unfortunately, Unity tends to break things often, so many of these tutorials are obsolete. But yes, it's a lot of tutorials and it's always possible to find the one you need.
226.
by krasin 3y ago
Wikipedia somehow handles this okay. But yes, there will be work to do to address the issue.
227.
by krasin 3y ago
Collecting training data is actually a perfect for crowdsourcing. Images/videos are easier than text, and text is easier than high-quality text, but all are doable.
228.
by krasin 3y ago
why?
229.
by krasin 3y ago
This is a case, where appeal to authority actually makes sense. Thanks, I accept your point.
230.
by krasin 3y ago
> If there is a problem put in heat set threaded inserts You would be surprised how many screws will go lose after a a thousand of hours of operation, even in this case.
231.
by krasin 3y ago
> I've used this technique on several of my projects. Did they have any moving parts? Did they experience continuous vibrations and frequent mechanical shocks? For DIY robot arms, fasteners are very often the issue #1, if arms opera
232.
by krasin 3y ago
My biggest concern is that they use screws over plastic as fasteners: https://github.com/PCrnjak/PAROL6-Desktop-robot-arm/blob/mai... Quote from the manual: > Screws are in this example M3 screws and holes
233.
by krasin 3y ago
The video of the robot arm in action: https://www.youtube.com/watch?v=PiBCbHtvbpE Very impressive repeatability and very smooth movement for a 3d printed arm.
234.
by krasin 3y ago
Anduril is a company founded by Palmer Luckey, also known for making Oculus VR. So, it's not that surprising that the same marketing style transitioned to the new place.
235.
by krasin 3y ago
> It’s to stop AWS from doing what they always do. Business Source License is a sane way to address that ([1], [2]). 1. https://mariadb.com/bsl-faq-adopting/ 2. https://fossa.com/blog/business-s
236.
by krasin 3y ago
The license is insane (custom taylored without a legal expert): https://huggingface.co/spaces/tiiuae/falcon-180b-license/blo... The gist is: > Commercial use: Falcon 180b can be commercially used but under
237.
by krasin 3y ago
JLCPCB PCB assembly service ([1]) is excellent and is really inexpensive. I used to reflow PCBs myself at home, but now I don't bother. 1. https://jlcpcb.com/capabilities/pcb-assembly-capabilities
238.
by krasin 3y ago
>The future of computing is that all code running on a device is one of the two S's: Signed or Sandboxed. >To do otherwise presents unnecessary risk. Unnecessary risk to whom? To monopolies that want to control the devices? I wou
239.
by krasin 3y ago
>If so, the 1.1 billion is not running costs, but capital spent to renew parts of the plant, replace some parts and so on That would explain it. Thanks for the data point. > Your point about solar is well made. That's actually so
240.
by krasin 3y ago
Yes. As much as I am a proponent of nuclear power, the proposed grant does not make sense to me - what exactly do they need this kind of money for? Operational expenses of properly built & operated nuclear plants are very low, which is

More ›