Personal favorite AI twitter follows, in no particular order: @nonmayorpete – Top tier AI reporting @sudu_cb – AI engineer @heyBarsee – GPT-5 the person @rowancheung – AI Twitter machine @Scobleizer – The Curator @jennyai – The fun AI news @DrJimFan – IRL smart person
@aibreakfast
-
The four stages of JARVIS workflow
By
–
The workflow of JARVIS consists of four stages: -Task Planning: Using ChatGPT to analyze the requests of users to understand their intention, and disassemble them into possible solvable tasks. -Model Selection: To solve the planned tasks, ChatGPT selects expert models hosted on
-
New Microsoft JARVIS with HuggingGPT and ChatGPT
By
–
New AI? A look inside Microsoft's JARVIS: Autonomous AI powered by HuggingGPT and ChatGPT (coming soon) GitHub: https://
github.com/microsoft/JARV
IS
… -
New book on AI for non-programmers
By
–
Today's newsletter includes a preview of our new book! Decoding AI: A Non-Technical Explanation of Artificial Intelligence is available April 18th. Fun book to write. AI 101 for the non-programmer.
-
Formal Algorithms for Transformers
By
–
9. Formal Algorithms for Transformers This document provides a precise overview of transformer architectures and algorithms, covering what they are, how they are trained, what they are used for, their key components, and a preview of prominent models. https:// arxiv.org/abs/2207.09238
-
DERA improves LLM responses in healthcare
By
–
10. DERA: Enhancing Large Language Model Completions with Dialog-Enabled Resolving Agents DERA is a framework that uses large language models to improve output in healthcare. DERA outperforms GPT-4 in medical conversation and care plan generation. https:// arxiv.org/abs/2303.17071
-
Language Models can Solve Computer Tasks
By
–
A computer agent that can understand natural language and perform tasks on a computer. They used a large pre-trained language model and a simple prompting scheme that helps the agent improve its output. https://arxiv.org/abs/2303.17491
-
ChatDoctor: A Medical Chat Model
By
–
A tailored large language model for the medical domain by fine-tuning it with doctor-patient conversations and information on diseases, symptoms, medical tests, and medications. Data made public https://arxiv.org/abs/2303.14070
-
ChatGPT surpasses human workers for annotation
By
–
—
4. ChatGPT Outperforms Crowd-Workers for Text-Annotation Tasks ChatGPT is more accurate, consistent, and cheaper than humans at organizing and labeling text, making it a challenger to MTurk. https://
arxiv.org/abs/2303.15056
— -
LLaMA-Adapter: Efficient Fine-tuning with Zero-init Attention
By
–
5. LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention LLaMA-Adapter is a fast way to improve an model's ability to follow instructions, producing quality responses and working with images for better problem-solving. https://
arxiv.org/abs/2303.16199