astle dsa

4.7K posts

astle dsa banner
astle dsa

astle dsa

@AstleDsa

Living in complexity, formalism, mathematics and computer science

Katılım Mart 2022
161 Takip Edilen1.1K Takipçiler
astle dsa
astle dsa@AstleDsa·
A Trillion Transactions by TigerBeetle Absolutely loved the talk + presentation. One of the few times a ~20 min video was over before I knew it. youtu.be/y2_BqkKTbD8?si…
YouTube video
YouTube
English
0
6
20
2.3K
judah
judah@joodalooped·
going to start making websites again (i took a couple weeks off)
English
1
0
20
682
astle dsa
astle dsa@AstleDsa·
@radokirov Although I'm not a mathematician by training, I always wondered if we could completely move to lean while doing math ! (scatterpad.com/page/01JYHRQF1…, some thoughts on the same) Do you think it's somehow limiting our imagination, to think in types, as compared to chalk and board ?
English
1
0
0
58
astle dsa
astle dsa@AstleDsa·
Rediscovering an old song I used to love was just a very good start to this day
English
2
0
4
419
astle dsa
astle dsa@AstleDsa·
Met a "mathematician" who had a bachelors in it, apparently liked math a lot and did not know what the millennium problems were lol I was in disbelief for straight 2 minutes, thinking he misunderstood me, cause no way he wouldn't know them, but he did in fact not know them.
English
0
0
0
497
astle dsa
astle dsa@AstleDsa·
@novasarc01 People be explaining the math, not realising that code is the best medium to convey these ideas for some reason
English
0
0
0
171
λux
λux@novasarc01·
a visual dive into conditional flow matching - this is an awesome iclr blogpost explaining normalizing flows, flow matching, conditional velocity fields...a must read if you want to deep dive into flow matching and physics behind diffusion models
λux tweet mediaλux tweet mediaλux tweet mediaλux tweet media
English
5
59
535
25.4K
astle dsa
astle dsa@AstleDsa·
@vxnuaj Feel like the the baity posts would be about agents, DeepSeek and the like
English
0
0
0
45
astle dsa
astle dsa@AstleDsa·
@wangbin579 The mathematical side of CS seems excessively hard for some reason. Especially proofs
English
0
0
3
132
wangbin579
wangbin579@wangbin579·
This is the hardest paper I've ever seen, and I still can't fully understand the proof. Math has no limit to its difficulty. If you can understand and use it well, you'll be truly outstanding.
wangbin579 tweet media
English
5
19
186
11.2K
astle dsa
astle dsa@AstleDsa·
@mu_chrinovic I think they should put out a list of their accomplishments while dissing on Linus lol He probably doesn't realise Git was made by Linus too, why not stop using that too ?
English
0
0
0
65
Chrinovic .M
Chrinovic .M@mu_chrinovic·
as each day passes I begin to hate this platform and everybody on it, especially tech part of it
Chrinovic .M tweet media
English
12
2
52
2.4K
astle dsa
astle dsa@AstleDsa·
@LivewJack Hello, I'm not aware of any web3 events happening around me (I've only attended one before). I'll always be interested in programming new ideas ! As I said, the technical aspect of crypto is still very fascinating.
English
1
0
1
16
Jack
Jack@LivewJack·
@AstleDsa Hey Astle! Was curious if you are planning on attending any Web3 Events this spring? Also, remembering your position on crypto, would you ever be enticed to build out one of your research ideas if there was funding involved?
English
1
0
0
16
astle dsa
astle dsa@AstleDsa·
It was kinda obvious but, shouldn't many of legal frauds and fraudulent practices take a hit, since any written content can be pushed into an LLM, and weed out the fraudulent part ? Seems obvious, but I don't think a lot of people/businesses are doing this, at least where I live
English
1
0
1
486
astle dsa
astle dsa@AstleDsa·
Open sets and closed sets not being mutually exclusive is so mathematics. Like a set can be open *and* closed ? Nice. Not confusing at all.
English
0
0
3
387
astle dsa
astle dsa@AstleDsa·
The most fun I have, programming, is: - Making games - Designing and making index.html sites (single file) - Writing C (fun because no dependencies. Want something? Code it.) - Implementing data structures (not as a part of any project)
English
2
1
46
1.2K
JT
JT@jiratickets·
@valhalla_dev don’t know a single one of those companies
English
4
0
18
1.5K
astle dsa
astle dsa@AstleDsa·
@joodalooped I treated mine like it was waterproof, paid the price
English
0
0
3
99
judah
judah@joodalooped·
i treat my Air like it’s made of rubber and so far it’s acted like it beautiful computer, always buy
Peli Grietzer@peligrietzer

New @Apple macbook airs have screens that get micro-cracks for no reason after 3 weeks of use and stop working and their warranty won't cover it. Learn from my experience and don't buy a new macbook air

English
2
1
47
1.8K
astle dsa
astle dsa@AstleDsa·
Looking at how DS has focused on everything from math, to hardware accelerations to file systems, no wonder they deliver Tbf many AI labs may be doing things at this level too, just closed source
DeepSeek@deepseek_ai

🚀 Day 5 of #OpenSourceWeek: 3FS, Thruster for All DeepSeek Data Access Fire-Flyer File System (3FS) - a parallel file system that utilizes the full bandwidth of modern SSDs and RDMA networks. ⚡ 6.6 TiB/s aggregate read throughput in a 180-node cluster ⚡ 3.66 TiB/min throughput on GraySort benchmark in a 25-node cluster ⚡ 40+ GiB/s peak throughput per client node for KVCache lookup 🧬 Disaggregated architecture with strong consistency semantics ✅ Training data preprocessing, dataset loading, checkpoint saving/reloading, embedding vector search & KVCache lookups for inference in V3/R1 📥 3FS → github.com/deepseek-ai/3FS ⛲ Smallpond - data processing framework on 3FS → github.com/deepseek-ai/sm…

English
0
0
3
369
astle dsa
astle dsa@AstleDsa·
@RoboFadi @pranavmarla Tbh AGI is kinda a very vague goal. I think the focus should be on making the most of current LLMs
English
0
0
0
29
Fadi "Ali"
Fadi "Ali"@RoboFadi·
@pranavmarla Valid point, scaling compute for inference might also face similar diminishing returns. We need innovative solutions beyond just scaling for AGI.
English
1
0
1
43
trees of thought
trees of thought@pranavmarla·
this is true if scaling pre-training is dead, can we really expect continuous returns on scaling inference time compute? remember, it's not guaranteed to get better like scaling pre-training if it does scale, can it not also run in diminishing returns like pre-training?
David Pfau@pfau

So, this is it then? No more OOMs of scaling, no more speculation about whether GPT-5, 6 or 7 would be AGI. The performance wasn't even good enough to brand it as GPT-5. The diminishing returns finally kicked in and this is as far as we'll get just from scaling pretraining.

English
2
0
4
438
astle dsa
astle dsa@AstleDsa·
While performing the exercise. Kinda weird too, because once I noticed it (cause it started hurting lol), avoiding it was as easy as simply consciously *not* doing it While drawing strength from different parts is good, doing so without being careful can do the opposite
English
0
0
2
107
astle dsa
astle dsa@AstleDsa·
Getting a leg cramp while doing pull ups seems ... weird, though perfectly normal when one experiences it (I was doing archer pull ups) I was unconsciously tightening my thigh muscles to get that extra drop of strength, which caused my leg to contract for longer periods ..
English
1
0
5
248