Room Temperature Superconductors, Llama2 and Single Photon Cameras – Live and Learn #22
Welcome to this edition of Live and Learn. This time it's a bit different because I've read less AI-related news than usual and instead found some interesting things related to biology and physics I wanted to share. As always, I hope you enjoy.
✨ Quote ✨
On our indeterminately long journey between now and death, we try our best to not wake up terrified every day of just how finite life is.
– Nat Eliason - (source)
Links
Room-Temperature-Superconductors. HOLY fucking shit. This might be the biggest news, even overshadowing all the AI hype this year. If these results are reproducible, a LOT of wonderful things will become possible. And reaching for the stars is going to become much easier since the technologies that room-temperature superconductors make possible are true sci-fi-magic. Even the people who wrote the paper realized as much and seem blown away by their own discovery, ending the paper with these words: "We believe that our new development will be a brand-new historical event that opens a new era for humankind." Damn, what a great time to be alive.
Cells are the Master Architects of Life. Most people think that genes control how our bodies are built. This, it turns out is wrong. Body layout is controlled by cells and not by genes. Cells are the true builders of our bodies, and genes "only" provide their toolboxes: proteins. But the architecture and locations of organs and limbs are determined by the interactions between cells. This article dives into how this works in more fascinating detail. On the same topic, you should check out the work of Michael Levin, who has worked on these ideas for a long time and did some amazing experiments showing how cells, in theory, can work together to build any sort of tissue we could imagine.
Llama 2 Release. Meta has released the next version of an LLM named Llama. This time it's officially open-sourced and was trained in cooperation with Microsoft. The model performs almost as well as GPT-4. However because it is completely open-source, you can get all the neural network weights and experiment and play with it on your own, no subscription to OpenAI needed. You only need the GPU compute to run it. I am very excited to see what systems people will build on top of this in the future and how this is going to accelerate progress in LLM research.
Protein Printers. This essay is about how we could approach designing and building a machine that could print proteins cheaply. The implications of such a machine would be huge, because proteins can do very useful chemical tasks, very efficiently. The problem is that even though we can predict the structures of proteins with machine learning models like AlphaFold, synthesizing them in bulk and experimenting with them in the real world is still a costly endeavor. With a Protein Printer, like the one proposed in this article, this cost would go down by a lot, which would unlock the use of proteins for everything.
Single Photon Cameras. This is another extremely cool thing I ran across in the last week. The paper this article is based on, found a way to build cameras sensitive to single photons. Cameras like this could in theory be used to look into the brain from the outside, with high resolution and in real-time. Which would be an incredible breakthrough in Neuroscience and a very valuable tool for understanding how our brains work. Right now their camera has 0.4 Megapixels, but as they keep working on it their resolution, and therefore the possibilities of the technology will increase dramatically.
🌌 Midjourney 🌌
🎶 Song 🎶
Four Fork Ache by José González
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.
Cheers,
– Rico