Conversation with Matt Dalio and expanding on the thoughts on a Generative AI for Engineering
By Lin Kayser
Published April 5, 2023

A few weeks ago, I found myself engaged in deep conversation with Matt Dalio, founder of Endless (and son of Ray Dalio).

With Endless, Matt is inspiring kids to actively participate in shaping the digital world, through coding.

When I learned how to write programs, more than 40 years ago, coding was the only thing you could do when you switched on the computer (you could start typing code before the CRT display had warmed up enough to show it…). Today, it has gotten so complicated to start, that many kids never do it, which is a shame, especially, because coding is now front and center, when it comes to the transformation from AI based models, to actual implementation. I have written about how we can use code as an intermediate step from a Generative Language Model to an AI that helps us engineer physical things.

Matt now wrote a great piece that also visually explores how such a world can look like, check it out here. (Thank you for the shoutout, Matt).

The most important part of it is to start encapsulating our engineering knowledge in computer code, that we can train AI models on. Ideally, there is a significant Open Source component to it, where engineers share their knowledge freely. If everything is locked behind corporate walls, we will have a hard time employing it to accelerate the pace of innovation.

One of the reasons, why Information Technology operates at the speed of Moore’s Law is, because in computer programming sharing source openly is quite common, if not the norm. ChatGPT was trained on millions of openly accessible GitHub repositories, and as a result can produce surprisingly complex functional code.

A conversational interface for engineering will depend on openly accessible encoded engineering knowledge.