Andrew Mayne is a writer, inventor and AI educator who hosts The OpenAI Podcast. He founded Interdimensional and is a general partner at Zero Shot Fund, following earlier work as OpenAI’s first prompt engineer and science communicator.[1][2]
Start with Why Tejal Patwardhan stopped underestimating the models - Episode 21 →
Education
Links
Website Books Interdimensional
X · AndrewMayne Founder · Interdimensional
General Partner · Zero Shot Fund
Host, The OpenAI Podcast · OpenAI
Career
Before OpenAI, Mayne worked in magic and entertainment, including the A&E series Don’t Trust Andrew Mayne and Discovery’s Ghost Diver. His biography connects the latter’s underwater stealth suit with work on shark models and a VR simulator. His later practice combines AI experimentation, teaching and writing fiction.[1]
Contributions
Explore capabilities and communicate the results
OpenAI’s GPT-4 contribution record names Mayne in novel capability discovery and communications. His later public work includes interviews with the people building models and products. In Episode 21, he asks frontier-evals lead Tejal Patwardhan about why old tests become too easy, how benchmarks can be gamed and what models should be measured on next.[6][4]
Use scientific inference as the engine of a thriller
The Naturalist follows computational biologist Theo Cray as he investigates a former student’s death. Where police suspect an animal attack, he sees evidence of a serial killer. The novel turns a scientist’s way of reading evidence into the mechanism of a mystery.[7]
Tools, writing & talks
Writing, research & talks
Start here
A conversation with Tejal Patwardhan about measuring increasingly capable AI systems.
Begin with the discussion of benchmarks becoming too easy, then consider which measurements match your own tasks. The original show notes introduce the questions; no model-training background is required.
Host and interviewer[4]Original source- The Naturalistandrewmaynebooks.com
A science-driven thriller following computational biologist Theo Cray.
Author[7] - GPT-4 contribution creditsopenai.com
The original record of contributors to model evaluation, capability discovery and communications.
Contributor: novel capability discovery and communications[6]