Parimarjan Negi

One questionable obsession to the next.

[Parimarjan Negi @parimarjan]

@GregHBurnham I have been working on a chess benchmark where experts would score ~80-100%, and most LLMs are < 50%. Not sure yet about the average good chess player (but will know more soon!)

[Parimarjan Negi @parimarjan]

2400 elo in bullshitting

[Parimarjan Negi @parimarjan]

@viditchess haha i wish

[Parimarjan Negi @parimarjan]

they fired everybody who couldn't figure out why kramnik's clock ran too fast https://t.co/Vwz5SUVnwj

[Parimarjan Negi @parimarjan]

@NikolaosNtirlis It is probably true that several people are cheating online, and statistical analysis may not help much in detecting it. also otb strengths are relevant. but weakest claim seems to be analyzing online streaks over short periods of time based on intuition from otb streaks

[Parimarjan Negi @parimarjan]

@NikolaosNtirlis I agree that outliers == cheating is too big a leap. And I meant anomaly detection at some intuitive level - I don't think the probabilities are valid. It is more like showing "obvious" proof that people might agree with and then putting up some numbers on top of it

[Parimarjan Negi @parimarjan]

@Nivedit75715980 @chess24com thanks glad to hear that, and yes my parents are

[Parimarjan Negi @parimarjan]

So is there no replacement for @chess24com candidates game recap articles that gives all the player opinions without forcing me to click on videos

[Parimarjan Negi @parimarjan]

@hartmannchess @chess24com @USChess I guess better than many others :-) But chess24 was great in how it got all the relevant quotes from twitter/video interviews/ other places in one place with the right context. Not quite easy to match...

[Parimarjan Negi @parimarjan]

@MichielAbeln @MarkTWIC twic is amazing for what it does, and has great breadth of events and no stupid videos, but the depth in the chess24 articles isn't really there in any other article sites

[Matei Zaharia @matei_zaharia]

At Databricks, we've built an awesome model training and tuning stack. We now used it to release DBRX, the best open source LLM on standard benchmarks to date, exceeding GPT-3.5 while running 2x faster than Llama-70B. https://t.co/QEx7gND6UJ

[Ajay @ajay9470]

Two back to back meetings with UROPs got me like

[Parimarjan Negi @parimarjan]

@jmftrindade this thing is useful

[Parimarjan Negi @parimarjan]

I guess they just ask chatgpt who to fire

[kirsten lum @kirsten_lum_]

I keep seeing, “The layoffs are in part due to greater use of AI.”

As someone who builds AI tools and helps companies adopt them, I just Don’t Buy It.

[Serafim Batzoglou @s_batzoglou]

I just realized I have Grok, was I missing something?

[Parimarjan Negi @parimarjan]

@SGradijan @MagnusCarlsen this local blitz tournament is getting a bit too tough recently 🥲

[Parimarjan Negi @parimarjan]

@HariChess just got fatter and old. you still do seem the same!

[Parimarjan Negi @parimarjan]

My 2023 PhD was great, I submitted 0 papers! Here's a thread on them. 1/1

[Parimarjan Negi @parimarjan]

@ali_smu 2024 has already been not so great, i postponed my submission to jan 1 just so i could make that joke

[Parimarjan Negi @parimarjan]

@agrimsingh happy new year agrim! may you accel even more haha