waiting for @untitled01ipynb to make a “if you use 100% of your parameters” meme whats the intuition behind the 10x size : 2x capability ratio? feels like even if the network is sparsely connected it should scale better than that
@swyx
-
AI Twitter needs less headlines, more hands-on expertise building
By
–
Too much of AI Twitter is nonstop headlines (what did {today's_character} see? who's on top of leaderboard today?). Not enough people just going heads down and building expertise*. I want to see more practitioner deep dives like @HamelHusain 's course: https://
maven.com/parlance-labs/
fine-tuning?utm_campaign=29ce77&utm_medium=partner&utm_source=instructor
… -
Feature Request: Summarization and Grammar Editing Discussion
By
–
yes that’d be great! currently only doing summarization and adding grammar editing. whats yiur usecae? lets dm
-
Curse of Knowledge: 1 to N vs 0 to 1 Thinking in ML
By
–
curse of knowledge and 1 to n mentality (minds go straight to optimization, reliability, cost, primarily respect scale) meanwhile the disruptive 0 to 1 stuff (capability, creativity, content) benefits from a totally different background classical ml folks dont have (so perhaps
-
OpenAI API New Chatty Toggle and Model Spec Framework
By
–
# New "Chatty" vs "No Yapping" toggle in OpenAI API and more upcoming I think @joannejang et al's new Model Spec is very thoughtfully designed and basically welcome it as the Three Laws of our times (in particular the Objectives/Rules/Defaults splits) But why is nobody talking
-
GPT-4 Regulatory Ringfencing and Latecomer Bias Critique
By
–
i thought it was fairly obviously drawn to ringfence GPT-4 (at 10e25) – always the classic regulatory bias of "what we have now is fine but no more without my approval ok" when being a latecomer regulator
-
Timeline Flooded with DBRX Content, Questioning Sponsorship
By
–
why is my entire timeline just dbrx propaganda rn did u guys buy twitter
-
Covering Data Sources and Topics with Pivot Analysis for AI News
By
–
ya and i basically cover all those although perhaps with some points of weakness. i need to do a "pivot" (as in pivot table) of the data sources with the topics so this helps with checking weighting. i think ultimately the dream is 1) "hacker news of ai" with these kinds of
-
Cursor Party: Competitive Programming with AI Enabled
By
–
i think either @amanrsanger or @arvid220u are speaking at AIEWF! cursor party?? could be fun to do some kind of competitive programming with cursor enabled
-
RAG for Diffusion Models to Reduce Knowledge Requirements
By
–
i wonder if we can do rag for diffusion models to get rid of the knowledge requirement. 0.6b params is great