I gave a talk at the MLSys conference in May this year, touching on various topics, including large-scale ML training systems, abstractions for embedding ML choices in computer systems, and CO2e emissions of language model training. https://
mlsys.org/virtual/2024/i
nvited-talk/2592
…
RESEARCH
-
Large-Scale ML Training Systems and Language Model CO2e Emissions
By
–
-
GPT-4 base is not aligned without SFT or RLHF
By
–
That’s why I qualified “aligned” — GPT-4 base isn’t aligned (i.e. no instruct SFT or RLHF).
-
AI and Sustainability Experts Shape Future at IIoT World
By
–
Meet the minds shaping the future of #AI and #sustainability! Don't miss insights from experts like Torey Penrod-Cambra, Josh Mattingly, and Sarvesh Saodekar at IIoT World Sustainability & AI Day 2024. Secure your spot today: https://
buff.ly/3YLVWsb #sponsored #IIoTWorldDay -
Clarifying what “aligned” means in AI industry
By
–
No, that is what it means. Philosophers of AI safety often mean something bigger, but in industry “aligned” just means “tuned for use” and implies some combination of SFT or RL.
-
Spatial Omics Reveals T Cells Driving Atrial Fibrillation Inflammation
By
–
An spatial omics discovery example from this week
In the fat tissue surrounding the heart, there are resident memory T cells that promote inflammation, prominent in people with atrial fibrillation, that helps explain vulnerability to this common arrhythmia https://
nature.com/articles/s4416
1-024-00532-x
… -
Python ML/MLOps CV/NLP LLMs AI Engineering Daily Tutorials
By
–
If you interested in: – Python – ML/MLOps – CV/NLP – LLMs – AI Engineering Find me → @akshay_pachaar Everyday, I share tutorials on above topics! Cheers!
-
Minimal Study Plan for Machine Learning
By
–
A minimal study plan for Machine Learning – Python
– Maths for ML
– Machine Learning
– Machine Learning Ops (MLOps) (links in next tweet) Check this out -
Experts Highlight Critical Gaps in Climate Action Efforts
By
–
Experts highlight the pressing gaps in #ClimateAction @wef #AMNC24 http://
ow.ly/cHja50Sr5Bs #ClimateChange #Nature #Decarbonization #EnergyTransition #AI #IoT #5G #Cloud @Hal_Good @gvalan @Analytics_699 @Xbond49 @JBarbosaPR @jeffkagan https://
crowdfireapp.com/video/b8a0fca0
f8db39f1d8ac36ffb4afac8a
… -
AI-powered coding: leveraging tests to enhance code quality
By
–
The main benefits of AI for coding are to be gained when there are automated tests in place? In this case, the LLM is exploiting the gap between the quality of the code (lower) and the quality of the tests (higher). It's like using computation to "pump" quality from the tests to
-
Iterative AI workflow for diverse story generation
By
–
Say you want kids’ stories meeting some criteria (length, topic). You ask GPT/Claude — stories are good, but low diversity and you need 10K. So: 1. Generate k examples from IF model
2. k-shot base model to get k+1
3. Validate k+1 for criteria via IF model