At Microsoft I met many. Always in committee. 🙂
AI
-
FTC lawyer frustration suggests regulatory action on tech
By
–
one FTC lawyer carrying a binder walks back to his seat looking frustrated
-
Data AI Summit 2026 Event Announcement
By
–
We're so excited! Make sure you're tuned in #DataAISummit
-
FTC vs Microsoft: closing statements strategy and legal representation
By
–
in these closing statements, four FTC lawyers are standing up and one Microsoft lawyer (Beth Wilkinson) is up. interesting approach
-
Waiting for the next training curve datapoint
By
–
that feeling when you are awaiting the next datapoint in your training curve to arrive
-
Pretrained Language Models: Looking Back at NLP Progress
By
–
I love looking back at our old @stateofai
… In 2019, we wrote that pretrained language models were making it big in NLP… "This method could be further scaled up to generate gains in NLP tasks and unlock many new commercial applications in the same way that transfer -
Major ML Conferences Adopt Research Track for Papers
By
–
Thanks for the kind words!! Over 30 papers used this track for ICML. NeurIPS and ICLR are committed to continue with the track, and ICML will decide after this year's conference.
-
First Machine Unlearning Challenge Launched by Researchers
By
–
Machine unlearning, an emerging #ML subfield, aims to remove the influence of a subset of training data from a trained model. To advance its development, we’ve teamed up with other researchers to organize the first Machine Unlearning Challenge. Read more →
https://
goo.gle/3JxVHsw -
OpenAssistant’s shift from RLHF to supervised instruction-finetuning
By
–
So on that note, OpenAssistant's earlier models used RLHF for instruction-finetuning (
https://
huggingface.co/OpenAssistant/
oasst-rlhf-2-llama-30b-7k-steps-xor
…); The later one seem to use supervised instruction-finetuning. Maybe @ykilcher has some insights whether RLHF wasn't worth the effort vs supervised?