How tiny can large language models get? For fun, I ran Llama 2 on a microcontroller. The GitHub repo shows how, plus some extra tricks.
EVERYDAY ROBOTS2016 – 2023
Everyday Robots was a moonshot at Google X to build learning robots that help people in their everyday lives. I led the engineering team making that possible with robotics and machine learning.
To learn more about the vision and work, listen to my interview on The Gradient Podcast, check out everydayrobots.com, or read pieces in Wired, The Verge, and The New York Times.
Much of the robot learning research now lives on at Google DeepMind.
An artificial intelligence that continues writing the Tractatus: Truth, poetry, and nonsense ex machina.
Play with it at wittgenstein.app, read the full story on Medium, and check out the code on GitHub.
“Artificial poetry is therefore one that does not arise out of consciousness, but lies in language itself.” So says Max Bense through GPT-3.
Read the essay about this philosophical experiment on Medium.
2022 – 2023
The Wittgenstein and Bense projects were
conceived around the same time and coincided with the release of GPT-3. Both play with the idea of continuing their writing through AI and explore emerging poetic and practical ideas.
The combined material eventually turned into a lecture series about language and AI at ToftH School in Berkeley.
CYPHER_WATCH is a brutalist wearable device built to manifest NFTs in the physical world, a Rolex for the cyberpunk future we are living in.
This genesis project of Inevitable Futures launched with an exhibit at High Line Nine galleries in New York. Head to Medium for more context.
Some of the watches are available on Foundation. Inevitable Futures is also on Twitter and Instagram.
What if you trained a GAN on the eBoy database? Get the story on Medium, NFTs on Foundation, and source code on GitHub.
Remember the Apple iSight? I resurrected the classic webcam by putting a cheap Raspberry Pi computer inside.
The Medium story explains how it works. GitHub has the open-source hardware and software. Tom’s Hardware and iMore wrote about it too.
It’s today’s New York Times print front page on the wall and nothing else. Because sometimes no user interface is the best user interface.
Learn about this build on Medium and read what Gizmodo had to say.
This pandemic project wanted to be a contactless fever thermometer with auto-aim.
Notably, It does this using thermal face detection on an Edge TPU.
Instructions and source code are on GitHub.
A little piece of calm technology: the smart picture frame with a pop of color and no cables.
Accent comes with a dynamic pixel city designed by eBoy. It’s also an art frame, calendar, and commute map.
Catch the why, what, and how on Medium. Code and CAD are on GitHub. You may want to follow @AccentInk for occasional updates.
A stock trading bot powered by Trump tweets. This is art, not investment advice. There’s an Elon version too now and it supports crypto.
I lay it all out on Medium and in a follow-up. Follow the original at @Trump2Cash and the other guy at @Musk2Cash. The code is on GitHub.
The New York Times, TechCrunch, and TheNextWeb had thoughts about it too.
Clean and simple smart displays that blend into the background. I still use these at home every day.
The first one was a bathroom mirror and I wrote about it on Medium. The Verge and Teen Vogue seemed to like it. I followed up with a table-top version and a making-of on Medium. TechCrunch covered it.
The software and hardware are open-sourced on GitHub.
2010 – 2015
The little face computer that used to be the future. Lightweight wearable technology that gets out of the way.
I was a founding team member and worked on software engineering and user experience design.
The infamous announcement is on YouTube. Somehow there’s still nothing comparable to Glass’ voice, navigation, and photography features today.
2007 – 2008
As a researcher at Osaka University, I worked on artificial muscle simulation in the lab of Professor Ishiguro (right).
There’s a whitepaper and some fun animations.
2007 – 2009
I co-founded this early augmented reality startup in Japan. See mentions in Communications of the ACM and Fast Company.
We were eventually acquired by Google. The technology pieces wouldn’t come together to ship until a decade later.