Paper reading club:
brew your own ChatGPT

Please read the papers prior to the seminar.

Following a recent "leaked Google AI strategy memo", we are exploring if OpenAI and the gang really do have a moat — by following recent work in small large language models and cheap finetuning.

When: 12 May Friday 19:00-21:30
Where: F0 main room (free entrance, donations welcome)

Question of the day:
Do scientists agree that you can brew your own ChatGPT at home and finetune it as you wish on your laptop in an evening? Do you need a stolen LLaMA for that?

Papers of the day:
- Chinchilla, the progenitor of small large language models
- LoRA, a finetuning technique at 1/10000 the cost
- Commoditize your complement, a principle of tech economics

We’ll walk through each paper and collectively resolve any unclear spots as we do so. Then we’ll share our perspectives on the subject and figure our how the paper fits into our life.

Shall you get stuck as you read, pop over to our chat — we’ll try to help.

cur470r @add512