Hyperreal Enterprises: Roadmap


This document synthesises a Roadmap (and perhaps also a Product Roadmap) for Hyperreal Enterprises, Ltd. The Roadmap is being written inside Org Roam (an Emacs package), and displayed with Firn. The document can be thought of as a Wiki (and edited inside Emacs). From these sources, other downstream formats can be derived. A narrative overview follows.

To skip it, browse ahead to Top.


We started thinking about how to create AI tools that would process open source information and enable applications such as automated computer programming. Then Open AI Codex and Github Copilot came along and slightly stole our thunder. But actually, these tools still leave plenty of open questions.

Focusing on the technical side of things: what would it take to build an AI system for programming, with depth measured not in terms of neural network layers but in terms of understanding? And what about other domains of inquiry, like science, mathematics, or collaborative design?

Here’s a curious thought. AI systems can beat the best human players at the game of Go, but these systems don’t “know” the basic concepts of Go that any amateur human player would recognise. These AI systems work with abstract patterns in data: in this case, patterns built out of three-by-three grids.

What if AI systems worked, instead, at the conceptual level? It’s possible to represent and reason about concepts on computers in a way that looks very different from word vectors. But the infrastructure for computing with concepts isn’t as well established as contemporary neural methods, and isn’t yet connected with those methods in a ready-to-use way. If we were able to bridge this gap, we could apply AI technology everywhere there are patterns.