Google presents an AI system to write expert-level scientific software.
Using LLMs + tree search, it invented novel methods in bioinformatics, epidemiology, geospatial analysis & more, often surpassing human SOTA. (1/4)
When you generate images with VQGAN + CLIP, the image quality dramatically improves if you add "unreal engine" to your prompt.
People are now calling this "unreal engine trick" lol
e.g. "the angel of air. unreal engine"
BloombergGPT: A Large Language Model for Finance
Presents BloombergGPT, a 50 billion parameter language model that is trained on a wide range of financial data.
arxiv.org/abs/2303.17564
Large Language Models are Zero-Shot Reasoners
Simply adding “Let’s think step by step” before each answer increases the accuracy on MultiArith from 17.7% to 78.7% and GSM8K from 10.4% to 40.7% with GPT-3.
arxiv.org/abs/2205.11916
We've released the weights (1.3B and 2.7B) of our replication of GPT-3 🥳
Using the updated Colab notebook in the repo you should be able to finetune the models on your own data as well as run inference.
Pattern of my 20s:
“This idea’s great, but others are better positioned. I’m late and lack domain expertise. I’ll find something new.”
→ Later: someone even less qualified makes it work.
If you think you have regrets, here are mine:
- I turned down an early invitation from Noam to join CharacterAI—and similarly from Igor to join XAI.
- I stumbled through a coding interview for OpenAI when Wojciech asked if I wanted to work on GPT-4.
- I was once trying to
