Joined August 2009
- Tweets 1,035
- Following 230
- Followers 211
- Likes 1,288
98 Photos and videos
Jan 10
Tonight's cocktail: kissel, or rather sbiten with lemon juice, cola syrup, honey and a splash of bourbon.
Slow-Motion Sour.
2
112
23 Dec 2025
Thought I am hallucinating that my computer freezes a bit when sun goes down 🧛
☞ Bug in Broadcom firmware caused freezing when jumping bands over load at dusk waves from weather radar bounce more from the sky, card perceive as congestion and searches for better band. Fixed 😊
1
51
4 Dec 2025
Not to shill for Claude Code (actually trying Antigravity now ;) but I love this ASCII art and on my terminal color scheme, this night with crescent Moon looks especially neat ^_^ Like, late late evening but not yet midnight, remains of cobalt color.
1
141
5 Jan 2025
Now I'm a man of culture too! I tried kuwai! I mean クワイ. I mean 慈姑. Tastes like a chestnut who became a root vegetable. Deep-fried. A bit bitter, but maybe that's from frying. Gives you that "find it in the forest and eat" feel, though this one was from the shop :)
3
101
Aleksandr Drozd retweeted
14 Oct 2024
📢 #NLProc I'm hiring! 2x PhD students (apply by Nov 22) & postdoc (Nov 15). These positions are part of a large project on the problem of attribution for LLM outputs to the original data sources, on which LLMs were trained.
Job links on my website (in the bio)
/1
3
49
175
31,478
Aleksandr Drozd retweeted
17 Jun 2024
Great news! Based on requests, the submission deadline for the LLMxHPC workshop (@ IEEE #Cluster) has been extended to June 29, 2024. Submit your novel research on using #HPC to advance #LLMs and using LLMs to improve the design/operation of HPC systems. llmhpc.github.io/2024/
1
2
177
arxiv.org/abs/2404.08676
🚨Introducing ALERT🚨 , a large-scale safety benchmark of > 45k instructions categorized using our new taxonomy.
✨Using prior work such as Bo Decoding Trust and our Aurora-m huggingface.co/datasets/auro…
and @jackclarkSF's awesome Anthropic-HH.
🙏
2
17
53
13,285
16 May 2024
Workshop on Large Language Models and HPC will be in Kobe Sept 24 at @IEEECluster, still more than a month to submit a paper! Organized by @kevin_a_brown , Tanwi Mallick, Juliane Mueller and yours truly. Details at llmhpc.github.io/2024, 🔁 appreciated!
2
7
1,304
9 Nov 2023
me: I’m pretty good with CSS. also me: and I don’t want to use JS on the server, so I’ll use a static site generator in Go...
Static site generator in Go: running a JS-based JIT-compiler for a CSS framework used by a theme I choose... JIT-compiling CSS... damn hipsters...
1
1
158
12 Mar 2023
As I'm trying to teach my daughter a bit of programming, had to brush the dust off the pycontextfree library... Here's the code for this one: github.com/undertherain/pyco…
1
2
189
Aleksandr Drozd retweeted
13 Feb 2023
#hpcmyths 6: We Will Soon Run at Zettascale!
Read our discussion in @ProfMatsuoka et al. arxiv.org/abs/2301.02432 and discuss here at the #HPC twitter!
Will we reach zettaflop/s performance or will fp64
lose relevance before?
38%
zettaflop in fp64 it is
39%
fp64 will not be relevant
23%
fp64 is irrelevant now
69 votes • Final results
1
3
6
2,385
18 Jan 2023
You can do animation in PDF slides with latex ONLY with multimedia package, ONLY with LuaLaTeX ONLY in okular (on linux).
JS based animation out for "security" reasons, XeLaTeX etc can't build multimedia, evince etc needs mouse click but in presentation mode it means next page 🤦
2
229
Aleksandr Drozd retweeted
9 Jan 2023
The 'Myths' paper is out: Satoshi Matsuoka, Jens Domke, Mohamed Wahib, Aleksandr Drozd, Torsten Hoefler. "Myths and Legends in High-Performance Computing" arxiv.org/abs/2301.02432" undergoing peer review. Hopefully the community will be engaged in yes/no discussions on Twitter!
17
110
361
123,031
14 Nov 2022
I think I have some sort of sleepwalking issue... I suspect I wake up at night, unconscious go to my PC and delete those email to which it was too bothersome to reply...
2
Aleksandr Drozd retweeted
25 Oct 2022
📢 #NLPaperAlert: Outliers Dimensions that Disrupt Transformers Are Driven by Frequency
with @gpuccetti92 @bkbrd Felice Dell'Orletta
TLDR: remember how BERT over-relies on just 48 magic params? They're related to token frequency in training! /1
arxiv.org/abs/2205.11380
4
5
39
28 Sep 2022
well, the #stablediffusion thing is indeed fun to play with. But feels rather then prompt engineering, might be better to shortcut directly to tuning the input vector.
1
30 Jun 2022
Doing #Anki for 14 years now 💪🏼. First was burn, then burnout, then finally steady pace. This is Japanese deck, but that's the only one I do regularly.
Days studied: 60% (3043 of 4996)
Total: 359690 reviews
Average reviews/day: 72.0
Average minutes/day: 9.3
1
4
26 May 2022
Compute enabled deep learning - are other AI paradigms waiting for their AlexNet moment? What are computational challenges? More insights from neuroscience? These and related topics at cortex.ws workshop by @APodobas, @barryjdevereux, @cdschuman, and yours truly.
4
10
Aleksandr Drozd retweeted
25 May 2022
.@aclmeeting will soon be over, but the fun continues. The Third Workshop on Insights From Negative Results is happening tomorrow! Here's what's on the menu:
insights-workshop.github.io/…
/1
1
5
34
18 May 2022
⭐Hey hey we are hiring! 🤖AI supercomputers awesome team experience of living in Tokyo, Japan 🍣🗻 ⛩️. Apply directly or ask me about the details. And please retweet! NLP, AI for science, compute platforms for AI - you name it.
riken.jp/en/careers/research…
2
31
79