Getting to 50% on the private test set on ARC-AGI will be easier than people think, and getting to 85% will be harder than people think. (Current high score: 39%.)
RESEARCH
-
Smart folks organize to address systemic frustration
By
–
There is enough smart folks here that are sick of this. We need a meeting of the minds and start organizing
-
Polygenic Risk Scores for Type 2 Diabetes Validation
By
–
It is high time that people know their polygenic risk score for actionable medical conditions that have been well validated across ancestries. Here's what a report should look like for Type 2 diabetes. https://
cell.com/ajhg/fulltext/
S0002-9297(24)00120-4
… -
LLMs as Memory-Based Answerers Without Reference Materials
By
–
The way to think about asking a factual question to an LLM is that it's a bit like asking a person who read about the topic previously, but they are not allowed to reference any material and have to answer just from memory. LLMs are a lot better at memorizing than humans, but the
-
OpenAI Acquires Rockset for Retrieval Technology Enhancement
By
–
Welcoming @RocksetCloud to OpenAI — great technology that will help power retrieval across our products and research, and even more importantly a great team:
-
Major Paper on Generative AI Economic Implications Published
By
–
This is one of the most important papers on the economic implications of Generative AI. It is great to see it published in @ScienceMagazine
. Congrats to @danielrock
, @manlikemishap
, @thankyouniceAI, and @sj_manning
. -
Methods for Preventing AI or Fighting Superintelligence
By
–
As a method for preventing AI, or a method for fighting superintelligence?
-
Cohere’s Aya-23 multilingual model performance review
By
–
https://
huggingface.co/spaces/CohereF
orAI/aya-23
… from @CohereForAI is supposed to excel in multi-lingual. Could you try and share what you think?