Any sufficiently advanced technology is indistinguishable from magic - Arthur C Clarke.
Nothing in my lifetime has this been more palpable than the moment the world met large language models late last year. Now, as the novelty has worn off and their limitations are discovered, the work of gradually iterating on them begins.
Alex and I originally met at Quora, where we worked on Quora Sessions together. We figured out how to effectively use experts’ and celebrities’ limited time to answer the public’s most in-demand questions. The safe and engaging platform we built was one of the first public social media channels that prominent figures began using. Yann LeCun, Hillary Clinton, and Yoshua Bengio all chose to use our platform for live writing sessions.
After our time together at Quora, I've watched in awe how Alex built Scale. I witnessed him truly accelerate the advancement of AI in the last decade.
During that time, I founded Insight, originally conceived as a search engine for doctors, which was powered by trusted data curated by expert clinicians and medical researchers. From there, it pivoted to become Insight Browser, the first extensible browser for iOS. Throughout the company's life, we sought to build tools that brought together the best of expert-level human intelligence with the best available AI systems.
Today, I'm excited to announce that I am joining Scale as Director of Product Design and Experts. My hope is to help achieve human-AI symbiosis that is prosperous, safe and most importantly, grounded in reality.
Building the Software Design and Development Teams of the Future
As AI models become more adept at generating ideas, searching knowledge bases, and executing actions via code, the discernment and wisdom of human experts grows in importance for assessing these models in terms of quality and safety, as well as for guiding the ongoing development of new generations of AI.
Interfacing human experts with advanced AI represents one of today's most pressing software challenges. As AI automates repetitive tasks in design and engineering, humans can focus on the thorniest problems of creativity and complex problem-solving. This interplay evokes two insights from George Pólya:
“If you can't solve a problem, then there is an easier problem you can solve: find it.”
"What is the difference between method and device? A method is a device which you used twice."
Experts tackle novel problems, devising solutions to encode as methods in AI models. Those models then assist in finding new solutions, fueling further progress.
Over the next decade, we’ll see this play out in every field of knowledge, but we’re seeing this very tangibly in software today.
-
How viable is chat as a universal interface for models?
-
To what extent can models architect systems and write code?
-
How can experts thoroughly verify a model for safety and accuracy?
-
What is the best way to gather human feedback to improve models?
These are challenges we’re grappling with today. Through partnerships with other leading AI companies and access to top experts, Scale is ideally positioned to breakthrough these problems. If this excites you, join our team! We're hiring across design, engineering, PM, and other roles to keep pioneering expert-AI collaboration.