chaxor

20 ms·

241.
by chaxor 3y ago
What about the heavy metals that are part of the cast iron and steel though? What about the heavy metals in ceramics? The list is pretty endless, but very specific types of glass (boron silicate or pyrex) and titanium can be good to switc
242.
by chaxor 3y ago
If only there was some model, something like a human, preferably smaller, that maybe doesn't live quite as long. Perhaps it would be ideal if we could even hold it in our hands, and have a small space to contain many of these models.
243.
by chaxor 3y ago
You can take a look at chinchilla or Llama scaling laws. It depends on what you want to optimize. If you're interested in inference optimal, go with llama, if it's compute optimal, go with chinchilla. GPT-4 technical paper also
244.
by chaxor 3y ago
This still doesn't make sense. P2P is web3? What does Blockchain have to do with it? Maybe git like Merkle trees can be involved a bit, which are quite similar to blockchains. I'm not understanding the hatred.
245.
by chaxor 3y ago
Don't advertise these languages, else they will become like the wretched python is now. Python used to be fantastic. It actually used to just work . That stopped once everyone caught on to the language, Nvidia got involved, and tenso
246.
by chaxor 3y ago
Any database easily handles terabytes, idk why that's a condition. DuckDB and SQLite easily handle terabytes of data. They're just not so much for cloud based apps, or things that need multiusers and access control junk. They&#x
247.
by chaxor 3y ago
My understanding of web3 is based on the idea of P2P first, and removing cloud based services as much as possible. Effectively torrent, IPFS, etc. I don't understand why people conflate currency (which HN despises) with that concept,
248.
by chaxor 3y ago
It's amazing how insanely slow OSes can be too install, even without providing some of these things. I feel like getting a new Debian docker image to apt update takes forever now compared to 1995.
249.
by chaxor 3y ago
You're on a forum for hackers, so most of the people here are the ones that care about the specifics of technology.
250.
by chaxor 3y ago
You're incorrect about better professors, but the extra money does allow for ease of research. The only reason research is perceived to be better at certain institutions is due to the extra money, which allows ease of research. Mos
251.
by chaxor 3y ago
This is very close to the academia/industry divide. Academia (the public) spends an enormous amount of money investing in research to discover something novel and effective, or makes a large efficiency gain. Industry reads that, and
252.
by chaxor 3y ago
They can increase the price further due to pressure. Problem solved.
253.
by chaxor 3y ago
I believe that Tock (tockos.org) and Theseus ( https://github.com/theseus-os/Theseus ) are in this area a bit as well, just from an actual OS perspective. I don't know much about this area, but it would be wonderful
254.
by chaxor 3y ago
If their statement is that they haven't changed the weights, there still an immense number of things they can toy with. For example, the prompt may add more 'safety' language in it, which can cause strange differences to occu
255.
by chaxor 3y ago
This seems more like a response from Bard or Llama. There is always some probability of any of these models responding in a nonsensical way like this however
256.
by chaxor 3y ago
I believe they are referring to the original observations of code output degradation due to adding training epochs for safety. These observations were made by Microsoft working directly with OpenAI early on. They comment on it in the techn
257.
by chaxor 3y ago
So people can view Twitter now? I thought they removed that feature not too long ago
258.
by chaxor 3y ago
Ballistic conduction means 'perfect' conduction in a sense. There's no resistance. It's only for armchair swcnts though. Other metallic swcnts are also great conductors, but if you separate out the chiralities (semico
259.
by chaxor 3y ago
It seems you didn't quite hear the argument. I agree with you that the models Google has released to the public are absolutely worthless. That certainly does not mean they don't have extremely performant models at all however. I
260.
by chaxor 3y ago
It doesn't look like anything to me. A lot of marketing, for sure. That's all that seems to crop up these days. After a few decent local models were released in March to April or so (Vicuna mostly) not much progress has really b
261.
by chaxor 3y ago
It would be nice to see 6 of them trained for different purposes by combining 5 of their outputs together and 1 trained to summarize for the most complete and correct output. If we are to trust the leaks about GPT-4, this may be a more fai
262.
by chaxor 3y ago
Google has far better models than llama based models. They just simply don't put them facing the public. It is pretty ridiculous that they essentially just set a marketing team with no programming experience to write Bard, but that sho
263.
by chaxor 3y ago
But they have to make up that data! Because the work they are doing now was based on a other trial where they made up data, so you have to fix this data to match what was expected from the previous studies. Of course we need to protect
264.
by chaxor 3y ago
This is important to point out. I was actually surprised by the mouthwash outcomes as well. Almost no one really talked about mouthwash, but it looked to be useful in the study.
265.
by chaxor 3y ago
This is the correct answer. Typically if something is done at an ivy league, other US and UK universities follow. Perhaps the other countries would follow as well shortly after that, or there would just be a divide between 'real&#x27
266.
by chaxor 3y ago
This is indeed fantastic, but paired with container proxies, it's awesome. You can send a few tabs through one socks proxy, a few through another, all while segmenting that from work, personal, etc. if you have a vpn for one work locat
267.
by chaxor 3y ago
Armchair swcnts are ballistic conductors. It matters quite a lot how that carbon is oriented.
268.
by chaxor 3y ago
This has already been shown to be true, many times. There are significant problems with swcnts (and even more so with dwcnts or mwcnts, due to less effective macrophage breakdown from h2o2 release). But there are some applications where the
269.
by chaxor 3y ago
You can't mention that on HN! You'll get aggro from all the 'anti-crypto' people! Clearly, this is an awful idea. Because something something web3 might be related, and something something 'crypto'! No clear an
270.
by chaxor 3y ago
This is just an absolutely awful idea simply due to the weight. The reason people are trying to get carbon nanotube electricity lines in the air is not because 'nano' is a cool word, but because they're really light while c

More ›