chaxor

36 ms·

331.
by chaxor 3y ago
Why on earth would this ever be accepted as a HN post.
332.
by chaxor 3y ago
It seems that most of these issues are not really a problem if the system is used for a small set of people on LAN only, with VPN if needed elsewhere. Perhaps the browser local storage is an issue, but may be mitigated if the credentials ar
333.
by chaxor 3y ago
Once you have an NN model and you know you want to keep the model weight set, moving to an optical neural network can massively drop the energy use. Its not necessarily easy, and certain architectures may not be as amenable, but it can cer
334.
by chaxor 3y ago
What is the correct tool to properly merge a large set of tar.gz files for which may have an enormous overlap of similar files, and some that have been altered just slightly? Git plus some parsing seems close in that space, as analyzing the
335.
by chaxor 3y ago
Sure, but this is unfortunately not really novel work in this area. White matter tract and (3T, typically not more) fMRI-based analyses of language tasks have been done for decades, and simply don't provide much information due to the
336.
by chaxor 3y ago
The main important part of open source, regardless of license, is that you have the ability to run it locally to ensure that the code written and provided is what is used for your interaction with the code (since obviously a company or en
337.
by chaxor 3y ago
Unfortunately this doesn't really work, as the model is not limited in it's decoding vocabulary. Does anyone have other suggestions that may work in this space?
338.
by chaxor 3y ago
Why on earth would you run ML on a potato?
339.
by chaxor 3y ago
Is there a decent way of converting to a structure with a very constrained vocabulary? For example, given some input text, converting it to something like {"OID-189": "QQID-378", "OID-478":"QQID-678"
340.
by chaxor 3y ago
This also doesn't make any sense, as it would be much cheaper to use local LLMs to generate text, making it likely to see generated toxic language.
341.
by chaxor 3y ago
This is only a good idea if you are *specifically not* using OpenAI. If you use local models then it's a fantastic idea.
342.
by chaxor 3y ago
... are robotics not at all even on the horizon of your thoughts process? The work from Voyager seems like an early PoC there already. Many hurdles of the past have been leaped over with ease lately in that area, so it's a glaring err
343.
by chaxor 3y ago
It's likely that it's because you have specific design principles in mind, whereas if you don't know about design in programming you will be blown away that you can copy/paste commands and something works. Otoh, if you w
344.
by chaxor 3y ago
The most useful part of the process is simply thinking about how to organize your files. The 03.65 like naming can indeed be switched out for something with words, but I believe the best of both worlds is to make the words "unix-like&q
345.
by chaxor 3y ago
This is a great point about tag maintenance *if you have to make the tags yourself*. However, if you have a simple ML system that you can run to categorize your files and pull out good single word descriptors that have a large explained va
346.
by chaxor 3y ago
I think it can be informative generally of what they may be targeting. Here, it seems that the interest is in somehow reducing the size and 'throwing away' some factual information for more reasoning capability (obviously this
347.
by chaxor 3y ago
Ok. Suggest a better theory so you can pat yourself on the back then.
348.
by chaxor 3y ago
How do you check it without running something that may execute it? Many scripts will execute if opened with vi or even cat or printf from what I understand.
349.
by chaxor 3y ago
Don't ever allow us by default. Use umatrix
350.
by chaxor 3y ago
I have a very different experience with GPT-4. It always provides about 6-10 references for every response, and they work perfectly every single time. I realize the hallucination problem and check everything every time, but compared to 3.
351.
by chaxor 3y ago
I think the point being made by using "Microsoft" here was basically a stand in for "other random people". I.e anything 'cloud'.
352.
by chaxor 3y ago
I don't know what is true either, but you can't run that analysis for the claim people are making. The reason being that people aren't claiming the API model outputs changed, they're claiming the website based outp
353.
by chaxor 3y ago
LLMs are useless! I was curious, so I ended up initializing one with 500 Billion parameters . I trained for a whole 4 hours on a whopping 100 books. It still doesn't know anything! Awful. Sad. Clearly, they can't reason. \s
354.
by chaxor 3y ago
Almost. Specific sections of neuroscience, which are actually just cell bio or biochem applied to brain tissue, sure. Often behavioral neuroscience has much of the same problems of not really having good definitions that map to tangible o
355.
by chaxor 3y ago
Wasn't there something called `teddit` that was written in nim that did a better job of removing all the js crap that makes reddit terrible? I would imagine everyone will just move to that if possible. Although perhaps that may also b
356.
by chaxor 3y ago
I think that's what was implied by "it's now or never". Your company forces anti-remote policies? If you're worth anything, just leave.
357.
by chaxor 3y ago
It's not really hard anymore at all. The installer gives you a TUI with options to select what you want, and pretty sane defaults. Setup is pretty easy and I would compare to the ease of Manjaro, although without the "snazzy gui&
358.
by chaxor 3y ago
Why not just use base Arch? It seems kind of silly to try anything else when the options are laid out very easily in front of you in the base installer.
359.
by chaxor 3y ago
Perhaps another example of us not really having a clue what's going on with biology, and how it's especially bad when this intersects with anything related with diet. Biology of aging is already extremely difficult, and that'
360.
by chaxor 3y ago
Probably better to skip that and go for characters or bytes, since it can simply learn morphemes or phonemes from the smallest structure available. Alas, the context size problem is the main pressure against this.

More ›