behnamoh

42 ms·

961.
Ollama vs. Llama.cpp(old.reddit.com)
3 points by behnamoh 2y ago|0 comments
962.
by behnamoh 2y ago
> It’s not a big deal if it remains so for another half a decade. Nobody outside of Apple has any expectations on it. But when Apple advertises literally all their new devices as "Built for Apple Intelligence" without actually
963.
by behnamoh 2y ago
Can someone explain why we care so much about graphs in various branches of science? Like, why model the problems mentioned in the video as graphs that need to be colored?
964.
by behnamoh 2y ago
jq has point-free programming. it's intuitive once you wrap your head around it. See this: https://en.wikipedia.org/wiki/Tacit_programming#jq
965.
by behnamoh 2y ago
the second pic reminded me of the pale blue dot... it's a long way from home...
966.
by behnamoh 2y ago
> Marginally, in some models only, and only in token generation, if I read the tables correctly. Which is exactly the point. If the benefits are so marginal, why not get a used M2 Ultra instead?
967.
M3 Ultra Is Slower Than M2 Ultra for Language Models(old.reddit.com)
3 points by behnamoh 2y ago|4 comments
968.
Ask HN: Best Setup for LLM training and inference with $30k?
1 points by behnamoh 2y ago|1 comments
969.
by behnamoh 2y ago
> We also change the architecture of the model to reduce the KV-cache memory that tends to ex plo de with long context This is key (pun not intended). It's one thing to run these models locally; it's a totally different game wh
970.
by behnamoh 2y ago
> The number of people using their phones in the hot tub, or in the sauna, astounds me on a regular basis. I can't imagine doing that. and the number of people who desperately look for a way to replace their batteries or upgrade to
971.
by behnamoh 2y ago
> They said it was to prolong battery life. it's like google saying they removed uBlock because their own approach is better for performance and the end user.
972.
by behnamoh 2y ago
no one asked for water resistance, but literally everyone said "we want replaceable batteries and don't care about phone thinness either", but Apple doesn't care.
973.
by behnamoh 2y ago
it is true, multiple sources reported that.
974.
by behnamoh 2y ago
I've read several books on history of science. in none of them did a scientist started with a philosophical idea which resulted in scientific discovery. it's always the opposite. e.g., quantum mechanics was the result of physics t
975.
by behnamoh 2y ago
That doesn't imply philosophy LED to scientific discoveries; rather it was a post-hoc consequence of said discoveries.
976.
by behnamoh 2y ago
this is the same team that a few months ago here on hacker news talked about how to do fine-tuning on large language models, and then made it close source.
977.
by behnamoh 2y ago
on the other hand, they might produce students that know how to use AI tools, and in the long-term that might be what businesses prefer – employees who know how to use AI, not employees who have memorized facts in school.
978.
by behnamoh 2y ago
that's just not true: https://www.bestbuy.com/site/apple-macbook-pro-14-laptop-m3-...
979.
by behnamoh 2y ago
if that were the case, then it would definitely help Apple Intelligence if the iPhone and Mac had higher amounts of RAM, but the base MacBook Pro announced by Apple a while ago had 8 GB of RAM and even the pro versions of the iPhone have 8G
980.
by behnamoh 2y ago
I said in parallel.
981.
by behnamoh 2y ago
related: https://aplaceofmind.notion.site/Tales-of-my-Tailscale-Usage...
982.
by behnamoh 2y ago
we’ve heard that claim for the past three years, but every effort by them points to the opposite. don’t get me wrong, I would love for Apple Intelligence to be smart enough on my iPhone and on my Mac, but honestly, the current version is a
983.
by behnamoh 2y ago
I know you’re referring to the exolabs app, but the t/s is really not that good. it uses thunderbolt instead of NVlink.
984.
by behnamoh 2y ago
> I can't wait to see someone testing the full DeepSeek model on this at 819 GB per second bandwidth, the experience would be terrible
985.
by behnamoh 2y ago
except that you cannot run multiple language models on Apple Silicon in parallel
986.
by behnamoh 2y ago
> My guess is that Apple developed this chip for their internal AI efforts what internal AI efforts? Apple Intelligence is bunkers, and Apple MLX framework remains a hobby project for Apple
987.
by behnamoh 2y ago
819GB/s bandwidth... what's the point of 512GB RAM for LLMs on this Mac Studio if the speed is painfully slow? it's as if Apple doesn't want to compete with Nvidia... this is really disappointing in a Mac Studio. FYI: M2
988.
by behnamoh 2y ago
but these AI tools let you customize the music according to your needs.
989.
by behnamoh 2y ago
whatever can be automated isn't "true" creativity. these models merely generate an average music, but the outputs of creative musicians always stand out.
990.
by behnamoh 2y ago
because copyright? these AI tools are great for adding a music to your YT video w/o worrying about violating copyright.

More ›