Do androids dream of AGI?

Last night I trained a CustomGPT on all my public writing. Having spent the holidays organizing two decades of blog posts I figured this half-megabyte of content should be a formidable starting point for creating a personal writing assistant.

Most of us have probably realized how limited and generic generative writing can sound without deliberate effort, and I suspected that preparing curated personal training data could significantly help on the self-automation front.

The first output (which isn’t this introduction) came out surprisingly decent. I submitted a draft article with feedback commentary, and asked GPT to rewrite it in my style. The piece ended up needing a new intro and light editing, but was 90% there within a minute. The subject matter was “known” by the GPT because it covered topics I have written about previously, but I was surprised at how good the response came out with such little effort. You evidently need to direct the production even after training to generate quality results. I’ll explore other sources of personal writing to supplement the data and better capture my tone and ideas.

This week’s issue features a fanciful selection of interviews and articles to help you make sense of the evolving landscape of GenAI and our inevitable progress towards AGI. While “general intelligence” means different things to different groups, it feels increasingly certain that the development of our tools and economies are moving in that general direction, and that staying informed to be the only sensible strategy.

A couple of readers showed interest in participating in a shared ChatGPT Teams account, but not nearly enough to justify launching a group just yet. I’m collecting use cases and building custom apps left & right and will keep sharing them with you here in future issues. LMK what resonates.

MZ

Excellent primer on the state of the art of AI when it comes to corporate innovation. Lux Research senior director Anthony Schiavo outlines all the different ways AI is being used in the innovation stack and outlines benefits and drawbacks which closely match my own experience building and experimenting with tools in this space. Highly recommended if you are in the innovation space, though note I disagree with some of his generalizations when it comes to how certain AI training data was generated. 🖇️

I’m a sucker for long-form interviews with tech CEOs because they are generally very telling of where the industry is heading. No notes, just a highly recommended conversation to listen to. 🏏

As someone who grew up playing video games in the 90’s, it’s difficult to accept that the NVidia has been carving out new forms of computation for 30 years, and how crucial they have become in the GPU wars.

Interview with the always wonderful Simon Willison.

Great thread via Tim Urban on Twitter.

YouTube interview between Dan Shipper and Geoffrey Litt. They talk about how ChatGPT can build working React applications and how this means you can build extremely niche applications that you woudn’t have considered working on before—then to demonstrate that idea, they collaborate to build a note-taking app to be used just during that specific episode recording, pasting React code from ChatGPT into Replit. Closely matches my own experience coding with ChatGPT and recently Copilot.