Скачать или смотреть THIS Will Make You Top 0.1% AI Researcher

Summary
The speaker argues that AI's future lies in processing raw bytes, not tokens, creating a unified model for all data. He documents his open-source attempt to build a "Byte Latent Transformer," a method that solves the long-sequence problem by grouping bytes based on their predictive difficulty. The project is coded with AI assistance and shared on GitHub for community collaboration.

Byte latent transformer - THIS Will Make You Top 0.1% AI Researcher — master the hidden frameworks elite researchers use to crack breakthroughs while everyone else chases trends. This one mental shift separates the pioneers from the pack — and it’s free.

arxiv - https://arxiv.org/pdf/2412.09871

GitHub - https://github.com/vukrosic/bitstream...

Full playlist for this research project - replacing tokens with bytes -    • Replacing Tokens With Bytes - AI Research ...  

Code LLM (GPT-5, DeepSeek, Qwen3, Llama4) from scratch - https://github.com/vukrosic/open-sour...

Novita, cheapest 5090 and 4090 (choose spot billing for 50% off) - use my affiliate to help me get more compute for videos, you pay the same - https://novita.ai/?ref=mjqyndm&utm_so...

0:00 - AI's Future: Raw Bytes, Not Tokens
0:28 - The Sequence Length Problem
1:04 - Grok's Rumored Byte Model
2:08 - A Call for Open Source
3:23 - The Next Big Things in AI
4:25 - Building a Simple Byte-Level Model
5:41 - The Computational Bottleneck
7:49 - Solution: Byte Latent Transformers
9:58 - How Entropy Patching Works
11:38 - Using AI to Code the Model
14:15 - The GitHub Project & Future Work -
-
Support my Free & Open AI research - https://ko-fi.com/vukai
Help me do this full time