Robotics, Rabbit R1, and the Beginning of the Post-Truth Era – Live and Learn #33
Welcome to this edition of Live and Learn. This time with crazy breakthroughs in the field of robotics, another LK-99-related possible superconductor, and people warning about the beginnings of the post-truth era, where deepfakes generated by AI are so easy to create it's not even funny anymore.
✨ Quote ✨
Exceptional people working together is the closest thing we have to a philosopher’s stone. Great teams are akin to true alchemy—creating something from nothing. Fight for them. It’s always worth it.
– Alex Wang - (source)
🖇️ Links 🖇️
Mobile Aloha by Stanford University. The demos for this have made the rounds on X in the last week or so because they essentially build robots that can mimic human tasks in a kitchen with surprising accuracy. They do so by learning through a tele-operation process. The crazy thing is that the whole model can run on a normal laptop with normally available hardware. This means that it is possible to build a robot like this on your own. The code is also open-sourced and can be found here on the Mobile Aloha Github.
Googles Advanced Robotics Project by Deepmind. Google is creating several spin-off papers based on its RT-2 architecture, aiming to train better robots in the future. Their work includes things like learning tasks from trajectories, ways to use vision and LLMs to essentially "auto train" new scenarios, and optimizing transformer architectures for use in robotics.
LK-99 reloaded? Another Possible Meissner Effect Paper. A university team from China found another possible room-temperature superconductor. This time around, the internet is less hyped and more sad to be let down again, yet there is also some hope and optimism going around. The prediction markets look grim though, with people collectively thinking that there is only a 3% chance for it being reproduced this time around. Let's see.
Post Truth Era by Ethan Mollick and Others. The video shows a deep fake of the creator and explains that demos like this can be easily created now. There's another post by Martin Nebelong an AI artist on X, that describes the same thing. They use tools like HeyGen and Dubbing for Everyone to create those videos and it freaks me out. The level of skill and source material required to make convincing deep fakes shrinks with every passing week (you need about 30 seconds of video material right now). Now it is impossible to tell real from fake at this point, especially if only casually looking at something while browsing the internet. As both Ethan Mollick and Martin Nebelong pointed out on their X accounts, now is the time to agree on a "passphrase" with loved ones, that can be used in correspondence to identify yourself. We live in interesting times and need to adapt the way we consume information from the internet in fundamental ways as the ratio of deepfakes to accurate information increases.
Rabbit R1 by Rabbit. This is a new device, that embraces LLMs and the possibilities they bring to create new and better user interfaces. Rabbit aims to fundamentally change how we interact with computation and create a revolution similar to the iPhone when it was first launched. The main idea behind Rabbit is that natural language and automatic fulfillment of complex tasks will become the new norm because of the power of LLMs and machine learning. And the Rabbit R1 is designed from the ground up with that in mind.
GPT Store by OpenAI. Since the OpenAI Dev Day last year, people have created over 3 million custom GPTs. Now OpenAI opened the GPT store so that people can start monetizing their custom GPTs and get financially rewarded for creating GPTs with high engagements.
🌌 Traveling 🌌
I've been sailing from Tangier in Morocco to Lanzarote in the Canaries during the last few days and had some of the coolest experiences of my life. Living on a boat and sailing is wonderful. And the sunrises and sunsets each day are truly magical.
🎶 Song 🎶
Swamp Thing by Hiatus Kaiyote
That's all for this time. I hope you found this newsletter useful, beautiful, or even both!
Have ideas for improving it? As always please let me know.