snowycat

The docket isn't available yet. The URL of the docket should be https://www.regulations.gov/docket/NTIA%E2%80%932023%E2%80%9... once it comes out, however.

The NTIA website says "Link to the Request for Comment is below. This will be updated with a link to the Federal Register once formally published, along with the final due date, which will be 30 days after formal publication and any other changes made in the final published version."

There are tools that you can use to force the model to give structured output, such as llama.cpp GBNF grammar for example. They're a bit harder to use than ask gpt4 but they do work pretty well for what I use it for.

You could always download it from github, turn on developer mode in chrome (there's a little switch in the top right corner of the chrome://extensions ui), and then load it as an unpacked extension from source without ever touching the chrome web store.

I am running 30b llama models (4 bit quantized using llama.cpp) on 32 gb of ram and no GPU. I get around 2 tokens/second.

It appears that the creator of the repo actually confirmed this: https://twitter.com/ekansa/status/1137052076062650368

In the diagram behind the tungsten penetrator, only the part of the penetrator that contacts the sabot has rings on it, while the rest of the projectile is drawn as smooth. The rings might have something to do with helping the sabot separate from the projectile correctly or preventing the projectile from slipping out of the sabot before it leaves the barrel.

The ADM-141 TALD is designed to be much cheaper than a HARM. It alone can't destroy the SAM, but it can distract it. I guess if you use enough of them you could call it saturation.

I fail to see how this is any different (for the purposes of getting around google) than any other VPN or proxy service out there. The proposed mechanism is just using a script that comes from the same server as the main website with perhaps slightly changed up code and a different file name to trick up adblockers. It can still fingerprint you without your actual ip address, as it collects data clientside.

Really impressive project! How did you learn all the things that you needed to make this? Are there any particular books or resources you would recommend? Also, how much did this entire project cost? I'd imagine processing that much data and storing it wasn't cheap.

This would mean that in order to send something back to that server you would need to send an equally powerful radio signal that could reach back to where the server is. Its not the most convenient thing for some people to have a high power radio transmitter in their backyard.

The first 5 digits are directly correlated to DOB and birthplace, the last 4 are mostly random. https://www.pnas.org/content/106/27/10975 There's a pretty interesting study done on this.

Asked a Russian friend of mine. The image of the page with lots of Russian text on it is about how to store magnetic tape.

That solar panel on the top seems awfully small to me. If you look at the solar panel on the helicopter that's already on mars, its almost 3 times bigger than the body. This just looks tiny in comparison.