@rishdotuk
Account based inIndia
About this account
- Account based in
- India
- Connected via
- India App Store
Account-level information from X, not a live location or the device used for a specific post.
Long document understanding, Multilingual Evals and efficient models mainly, but other #NLProc applications in free time | vim enthusiast
Joined April 2011
- Tweets26.5K
- Following1.5K
- Followers890
- Likes78.1K
Thank you @kunalb11 for this masterstroke to remove that @WhatsApp spamming of messages from every business without explicit consent.
Rishu Kumar retweeted
So yeah how is nobody in jail for all these AI cyber incidents? Aaron Schwartz was prosecuted to death for like 0.0001% of this.
Rishu Kumar retweeted
Who wants more speech training data?
YODAS v3 is now available @huggingface
It’s 1.1M hours - the biggest audio dataset ever. And the first at this scale with stereo audio at 48kHz, with timestamped transcripts and translations. 100+ langs
hf.co/blog/espnet/yodasv3
CC-BY-3.0
Rishu Kumar retweeted
Apparently you can now outsource your research experience, recommendation letters, and academic references for $3,000.
I received the proposal today. 😭
Rishu Kumar retweeted
Exactly this, it's not only not healthy but it's questionable science. Even with fast pace it's so important to stop and think about whether what you are doing even make sense (don't do things just because you can)
Gonna go to the samsung service center and have my humiliation ritual of paying someone to replace the battery on my phone when I can do it myself way quicker if they sold me the freaking battery.
the gemma 4 12b is a beautiful/underappreciated model man... fully linear / standardized multimodal embeddings... reasonably deep... not so large where being dense vs MoE starts to be questionable (i.e 30b)...
say what you want about google's RL, but their pretrains are elite
hey guys! look at this extremely useless and buggy thing my claude cooked up in just a few prompts!
i haven't looked at the code! its in rust though! it's fast for sure, im going to tweet about this now with a link to my github because nobody else must've done this before
Rishu Kumar retweeted
What BigToken doesn’t want you to know is that they also read arxiv and X to get new ideas. Models are good at magnitude, direction is still hard. There is some uncertainty around for however long though, but its not today
Please continue the amazing works like looping transformers and echo.
Rishu Kumar retweeted
Yep. This is a toxic (and crazy) outlook. There are plenty of important problems that basic research will solve outside of labs.
Much of this pessimism btw is created by people that have never experienced excellence in pure research and thus don't have a frame of reference.
Rishu Kumar retweeted
One of the worst forms of brainrot AI has cultivated is pessimism about basic research, ie the idea that important work can only happen inside a frontier/neo lab and only with 10k+ GPUs, so the rest should not even bother. What a bleak way to think about science. And it's false.
Rishu Kumar retweeted
I see people being sarcastic but there is a real demand for it: if you want to embed models in industrial processes you need to ease audit and doing local language (and even more local professional register) remove frictions.
It is a tool to review/make things. MFs act as if they are debugging cuda-graphs everytime they are using an LLM, while the best they end up doing was to center the div on a slop website or write a stupid python script which is painfully so verbose that will make Guido screech.
Rishu Kumar retweeted
Uh yeah what did you guys think LLM as a judge meant
BANGALORE DISTRICT COURT POSTED THE CHATGPT PROMPT IN THEIR JUDGMENT 😭😭 SON😭😭😭 indiankanoon.org/doc/1955248…
