I’m happy to share my first open-source library:
Autonomais – a typescript library created to streamline the coordination of autonomous agents.
Designed with simplicity and efficiency in mind, Autonomais helps developers build AI agent workflows.
Built on LangGraph.
Oner is an AI-aware notebook, built on the idea that agents should be an integral part of every day creative work. Swarms are AI agents that can cooperate to accomplish a common goal or purpose.
See Oner here: https://www.oner.chat.
This is a simple idea…
Today, we have custom instructions in ChatGPT, that we can use to define how the chat AI interacts with us.
Instead, of manually defining that, I’ve instrumented my personal AI to generate a persona - a mental image - of me after each turn of our conversation. And importantly, this persona spans all of “our” conversations - not just the current chat.
Below is an example. Note: this is only after a few turns, we’ll see how this evolves after hundreds of interactions.
I recently improved how memory is managed in my work on interactive AI agents, and am pleased with the results.
This is a great read on some of the considerations of microservices (and a software approach to addressing).
Two research papers this year demonstrate that generative AI can significantly enhance productivity in the workplace.
Today’s LLMs are pre-trained, then frozen in time (unless fine-tuned). What changes when they can continuously learn?
I’ve been having an absolute blast with AI (GPT3.5/GPT4 from OpenAI) and wanted to share some of my enthusiasm.
Early versions of GPT-4 could “solve novel and difficult tasks that span mathematics, coding, vision, medicine, law, psychology, and more, without needing any special prompting.” - Sparks of Artificial General Intelligence research paper by Microsoft Research, April 13 2023.
The stunning advances in AI this year further underlines the imperative that every company needs to invest in incorporating these capabilities into their business.
The United Nations recently estimated that the world’s population had surpassed 8 billion people. That number has doubled in the past 50 years.
The beginning of AI could be characterized as sophisticated pattern recognition. ML can identify patterns in data, such as classifying whether an image is of a dog or a cat or if a radiograph contains cancer. This is extremely valuable for a broad class of problems.
A wonderful gift and a great example of the kind and generous people that they were.
Big update from Meta on an advancement to a general purpose learning algorithm:
There are parallels to Sam Altman’s recent essay, “Moore’s Law for Everything”, with the well-known essay by Marc Adreessen, “Why Software is Eating the World”, shared in 2011.
Perhaps in a decade, we’ll look back with reverence on this latest essay in the same way we view the 2011 entrant. With the continued advances in AI, it does seem like he has it right.
Over the weekend, my good friend and mentor, Steve Nordberg, passed away.
I recently joined Structural (structural.com), the Employee Success Platform. Moving to a startup has long been a dream of mine. But, the right one needed to be… well, right. So, what follows are a few reasons why I chose Structural.
To my colleagues at VUE,
This past Friday was my last day at Pearson VUE - marking the end of over two decades.