Simulators :: — Moire
Bot credits Model GPT-3 Tokens 1,455 Gain (bits) ρ (bits/token) λselection (tokens/bit) Selection 196.82 0.14 7.39 bold is prompt Outline: Recently, a relatively unprecedented type of AI has emerged, achieving (and totally transforming) SOTA on diverse tasks, and dominating researcher attention Massive (transformer) models trained on massive, self-supervised datasets Exemplified by GPT/codex “Foundation models” Referring to the tendency for these pretrained models to be used as the “foundation” for further fine tuning/RL But that is a prescriptive, not descriptive category Motivation: GPT is the most powerful AI we’ve created and it works very differently than others (real and hypothetical) How does it interact with preexisting and hypothetical types of AI?
Simulators Bot credits Model GPT-3 Tokens 1,455 Gain (bits) ρ (bits/token) λ selection (tokens/bit) Selection 196.82 0.14 7.39 bold is prompt Outline: Recently, a relatively unprecedented type of AI has emerged, achieving (and totally transforming) SOTA on diverse tasks, and dominating researcher attention Massive (transformer) models trained on massive, self-supervised datasets Exemplified by GPT/codex “Foundation models” Referring to the tendency for these pretrained models to be used as the “foundation” for further fine tuning/RL But that is a prescriptive, not descriptive category Motivati
related reading
- Simulators — LessWronglesswrong.com
- gpt-4.pdfcdn.openai.com
- GPT-4openai.com
- The Scaling Hypothesis · Gwern.netgwern.net
- Janus' Simulatorsastralcodexten.substack.com
- interpreting GPT: the logit lens — LessWronglesswrong.com
- The upcoming GPT-3 moment for RL | Mechanize, Inc.mechanize.work
- GPT in 60 Lines of NumPy | Jay Modyjaykmody.com
- Cyborgism — LessWronglesswrong.com
- GPTs are Predictors, not Imitators — AI Alignment Forumalignmentforum.org
- Scaling: The State of Play in AIoneusefulthing.org
- GitHub - brexhq/prompt-engineering: Tips and tricks for working with Large Language Models like OpenAI's GPT-4.github.com