Aran Komatsuzaki
Poisoning Web-Scale Training Datasets is Practical

Shows how to effectively poison 0.01% of datasets like LAION-400M for just $60 USD and degrade model performance.

https://t.co/qO420ZjxZh https://t.co/HfPecNPqlk
tweet

proxima centauri b
GN 🌚 https://t.co/5XcskVwyUG
tweet

AK
How Good Are GPT Models at Machine Translation? A Comprehensive Evaluation

abs: https://t.co/jREoE9Iiuh https://t.co/u2U6b7XRVW
tweet

Dave Craige
RT @alexandr_wang: one thing that most AI hypers don’t appreciate is just how slowly most of the world changes

most workloads at big banks still are 80%+ on-prem, despite the cloud existing for 16+ years

true AI-everywhere adoption will take decades, despite how mind-blowing the tech is
tweet

AK
Cross-domain Compositing with Pretrained Diffusion Models

abs: https://t.co/pexSKb1aAf https://t.co/FQy9fsG1Fg
tweet

Mike Tamir, PhD
Learning Observable / Observable | Observable https://t.co/YhxXxWfVAr
#AI #MachineLearning #DeepLearning #DataScience https://t.co/AUr7XGwF9m
tweet

Dave Craige
What is the most popular crypto company in the Philippines?

Which one should my VA’s start with first?

So she can easily transfer from crypto to Philippine peso (PHP)?
tweet

Earth Is A Sales Funnel For SATAN
expecting reactions without actions - and inaction without consequences - not how it works
tweet

AK
NerfDiff: Single-image View Synthesis with NeRF-guided Distillation from 3D-aware Diffusion

abs: https://t.co/xajK2VrcdC
project page: https://t.co/KtVopBjVwW https://t.co/QfmPDoOiAB
tweet

Robin Hanson
I suspect some need a reminder of the simple logic of free speech: if you ask me to force others who don't trust you, even though you could have earned their trust, to act as if they trust you, my default is "no".
https://t.co/HyHoLU5g8U https://t.co/zLa0IZz3oA
tweet

Rive
RT @AlexDelker: Sketchin'! @rive_app https://t.co/ZoRkVDAJVt
tweet