It’s crazy in hindsight that the stylistic mimicry of the GPT-2/3 era was such a local maximum for AI-written texts as cultural artifacts. For all their gains in making sense, aligned LLMs still can’t write with the eerie naturalism GPT-3 exhibited four years ago:
RESEARCH
-
Questioning model language understanding versus ‘4o’
By
–
What kinds of language understanding is it struggling with? Also, this is relative to 4o?
-
Grok-2-mini Speed Upgrade: Inference Stack Improvements
By
–
Grok-2-mini just got a speed upgrade. Over the past few days, we have substantially improved our inference stack. These gains come from using custom algorithms for computation and communication kernels, along with more efficient batch scheduling and quantization.
— xAI (@xai) 24 août 2024
Our inference… https://t.co/pwAG5EiMNAGrok-2-mini just got a speed upgrade. Over the past few days, we have substantially improved our inference stack. These gains come from using custom algorithms for computation and communication kernels, along with more efficient batch scheduling and quantization. Our inference
-
Knowledge Beats Intelligence: The Cost of Applying vs Solving
By
–
Gathering and maintaining knowledge is expensive and requires intelligence. Applying it is vastly cheaper than using intelligence to solve the problem. We've learned this in AI over and over again. And it's true for humans as well: being intelligent is no match for knowing the
-
AI Scientist: Automated Pipeline for End-to-End Scientific Paper Generation
By
–
New from @SakanaAILabs
! The AI Scientist Can AI be used to write high-quality scientific papers? The AI Scientist offers a fully automated pipeline for end-to-end paper generation. Leave your thoughts here for authors @_chris_lu_ @cong_ml @RobertTLange to respond! -
Master Your Craft: Expertise Beyond ChatGPT Capabilities
By
–
Master your craft so well that ChatGPT begs you for expertise.
-
Deliberative participatory AI design approaches and governance frameworks
By
–
Great letter! Deliberative and participatory AI will need many forms, and Jack brings together the concerns excellently. We had an approach called STELA relevant to the design role of deliberation 1/2. https://
nature.com/articles/s4159
8-024-56648-4
…
