according to my napkin math, about 1 billion people use BPE tokenizers (/LLMs)
RESEARCH
-
Data Augmentation vs Generative Models: Key Differences Explained
By
–
yes, a kind of flavor of it. but also not quite. in traditional data augmentation (say image/cnn contexts) you add noise to the data, rescale it etc. here, the difference is that you actually generate that data. in image/cnn land that would be using a diffusion model to generate
-
Linear Approximations in Modern Machine Learning: Why Simplicity Works
By
–
that's nowadays literally a multi-million dollar question . Btw there are lots of (linear) approximations that came and went in the past few years, but it's still more or less the same olde because it just works so well!
-
AI More Than Fancy Autocomplete: Investment Justification Debate
By
–
‘You’re making a lot of assumptions that no one can predict’ says the guy that’s making the (massive) assumption that AI is fancy autocomplete that doesn’t warrant the investment (and presumably r&d and innovation?) it’s commanding from smartest, most successful people in the
-
Autocorrect Technology Demands Half Trillion Investment Awards
By
–
‘Functionally advanced autocorrect’ warrants a $500bn investment, wins two noble prizes in a year and solves one of the most complex challenges in biology then, eh?
-
China AI Compute Model Challenges US Training Infrastructure Scale
By
–
Your point is badly conceived. Just because China built a model that requires less compute than the US ones doesn’t mean thar compute doesn’t need to be scaled. The compute isn’t built for model training long-term. It’s for inference and real-world use cases.
-

ByteDance Releases Doubao-1.5-Pro AI Model
By
–
ByteDance Doubao-1.5-pro is here https://
team.doubao.com/en/special/dou
bao_1_5_pro
… -

DeepSeek-R1 AI Model Released on Hugging Face
By
–
DeepSeek-R1 is here https://
huggingface.co/deepseek-ai?co
ntinueFlag=f18057c998f54575cb0608a591c993fb
… https://
github.com/deepseek-ai/De
epSeek-R1/blob/main/DeepSeek_R1.pdf
… -

Kimi K1.5: Scaling Reinforcement Learning with LLMs
By
–
Oh!!! kimi k1.5 is coming! Kimi k1.5: Scaling Reinforcement Learning with LLMs https://
github.com/MoonshotAI/kim
i-k1.5?tab=readme-ov-file
… -

OS-Genesis: Automating GUI Agent Trajectory via Reverse Synthesis
By
–
OS-Genesis: Automating GUI Agent Trajectory Construction via Reverse Task Synthesis https://
qiushisun.github.io/OS-Genesis-Hom
e/
…