novaRom

9 ms·

61.
by novaRom 2y ago
> stops all notifications after 20:00 Better to stop absolutely all notifications. When I need to know if someone of my friends is asking something I can open a chat app and check. If something is urgent then just a phone call. The amoun
62.
by novaRom 2y ago
What Pi 5 is good for? It doesn't look like a portable because of high power consumption, nor it's a desktop class system because of its weak compute and high price. Pi 3 or 4 is still a better choice for almost anything including
63.
by novaRom 2y ago
Without sarcasm, I’ve never been able to understand how or why real people use Twitter or similar platforms. Do actual human users genuinely scroll through and read all that noise? What motivates someone to do that? Is it comparable to the
64.
by novaRom 2y ago
They are really helpful to answer concrete questions, basically a replacement of manual web search and filtering, getting just exactly answers I was looking for. For example, one of my dialogue with chatgpt 4o was about boosting plant growt
65.
by novaRom 2y ago
This strategy is not unique to one particular country. Basically, it happens on different scales but repeatedly in the history (like Japan post-war).
66.
by novaRom 2y ago
Is German Economy in a free fall at the moment? Looks like many companies are in a "zombie state" unable to survive high interest rates, high inflation, and widespread job losses happen mostly among high-income individuals? "
67.
by novaRom 2y ago
Search Engines will probably not disappear in the near future even if most information will be processed via LLMs. I wonder if certain open (meaning not via Google/Bing) data processing is possible just to have fully local LLM with up-
68.
Ask HN: If Local LLMs Are Possible, Why Not Local Web Search?
3 points by novaRom 2y ago|4 comments
69.
by novaRom 2y ago
I am a bit skeptical. There were a lot of papers and experiments in 80s and 90s about different ANN architectures alternative to f(x*w+b). The reality today all practical SOTA models are still multiply-accumulate-threshold based. It's
70.
by novaRom 2y ago
Agree in general, although cheese (Old Amsterdam) and flowers (tulips) are fine for their price.
71.
by novaRom 2y ago
But tomatoes from Spain and Morocco taste much better than those from Netherlands even if transport is significantly longer - how to explain it?
72.
by novaRom 2y ago
A generator of "all primes" is pretty simple and deterministic. But you cannot simply generate next prime given only prime n without recomputing its non-trivial remainders. That means just a binary representation of a number n doe
73.
by novaRom 2y ago
Just kernel drivers. To know what is inside, you can disassemble them with https://github.com/NationalSecurityAgency/ghidra
74.
by novaRom 2y ago
Chances if Microsoft or Crowdstrike will be held liable for financial losses caused by this outage?
75.
by novaRom 2y ago
A perfect place for multimodal Assistent would be in Operating System's GUI. So that it sees and hears the same as its human user.
76.
by novaRom 2y ago
I think similar approaches might be useful in micro-electromechanical systems ( https://en.wikipedia.org/wiki/MEMS )
77.
by novaRom 2y ago
Probably not. There will be better GPUs. It's like we did use all those Kepler K10 and K80 fifteen or so years ago, they were Ok for models with few millions of parameters, then Pascal and Volta arrived ten years ago with massive speed
78.
by novaRom 2y ago
We are entering new era of building new competitive things "from scratch" and building them really fast. Powered by LLMs, increased personal productivity, ease to access knowledge, it's just inevitable a lot of better things
79.
by novaRom 2y ago
Interestingly, this article itself is an example of "slop"
80.
by novaRom 2y ago
> my personal ideal weather _is_ cool/gray/wet with ideal outside temperature of 15C 15C day temps? morning/night temps can be much lower. Winters in Norway are much colder. If you are looking for stable day-night temps ye
81.
by novaRom 2y ago
> more efficient than standard attention whenever the model dimension is greater than or equal to the context length all practical models have context length significantly larger than model dimension
82.
by novaRom 2y ago
Why did they build them? Any practical advantage to live so high?
83.
by novaRom 2y ago
One important element is to provide a workaround (info, instructions, or a small script) ASAP before you start fixing it, especially important for all kinds of blocking issues.
84.
by novaRom 2y ago
I wish we had a theory connecting randomness with time/space complexity. Intuitively I think the key to many computational limitations is making use of randomness.
85.
by novaRom 2y ago
> beauty of the GPT-based LLMs is their structural simplicity human brain's structure is also encoded in a short DNA sequence
86.
by novaRom 2y ago
In AI music generation we have much better results with large vocabulary sizes of 10^6 order, my uneducated guess is that's because transformers are not universal pattern recognizers, they can catch patterns on a certain granularity le
87.
by novaRom 2y ago
There are multiple layers where interception can happen: 1) On-screen keyboard - by default most phones do send what is being typed - a lot of phones also have 3rd party keyboards of doubtful origin preinstalled 2) "Enable backup"
88.
by novaRom 2y ago
Quantum effects are everywhere: https://en.wikipedia.org/wiki/Quantum_biology
89.
by novaRom 2y ago
Also hard vs soft water may sound different, not just carbonated water. Soft water feels smoother and silkier on the skin because it contains less minerals.
90.
by novaRom 2y ago
I have asked chatGPT to write a hokku: In garden's silent nook, Beneath cherry's tranquil look, Blades retire from the cook.

More ›