Talking Robots, Futurehouse and the Role of Computation & AI in Biotech – Live and Learn #28
Welcome to this edition of Live and Learn. This time with a company trying to build semi-autonomous AI researchers, a TED talk by Stephen Wolfram on Computation, AI, and the nature of the Universe, a talking robot tour guide by Boston Dynamics, a dataset designed to help algorithms learn emotional intelligence, and more. As always I hope you enjoy this edition of Live and Learn.
✨ Quote ✨
Surprises are things that you not only didn't know, but that contradict things you thought you knew. And so they're the most valuable sort of fact you can get. They're like a food that's not merely healthy, but counteracts the unhealthy effects of things you've already eaten.
— Paul Graham - (source)
Links
Trying to Build Semi-Autonomous AI Researchers by FutureHouse. This non-profit research institute is aiming to automate bioscience research within the next 10 years. They believe that the current advances in AI are already enough to build a useful tool that can help formulate and answer research questions semi-autonomously. The plan is to increase the number of hypotheses researchers can generate and test by 10x or 100x, thereby drastically increasing the capabilities of research as a whole. It's refreshing to see a company that is organized to chase long-range "moonshot" goals, trying to help humanity. And the backing (from people like Eric Schmidt) as well as the founding team sound like an impressive list of people so I am excited to see where this is going.
Making Chat (ro)Bots by Boston Dynamics. Boston Dynamics is putting ChatGPT paired with voice generation and recognition into their robots to allow them to walk around and converse with people, as some sort of robo tour guide. It's amazing how relatable a robot like Spot becomes once you give it a glued-on mustache, some Googley Eyes, and a little hat. Human brains really want to see faces everywhere and relate to machines emotionally and anthropomorphize them. Still, these are some pretty cool developments, and enhancing robots with solid real-world, real-time language capabilities is a big step forward.
Open Empathic by Laion. The group behind the datasets used in training Stable Diffusion is working on building a new dataset for characterizing human emotions. Detecting and responding accurately to human emotions is something that AIs can still improve on, and hence we need better training datasets, they argue. Especially for AIs that need to interact with humans in the emotional realm (think therapists or teachers) a dataset like this can come in very handy. Because with it we can teach an AI emotional intelligence, which is a big deal. This might lead to some pretty interesting applications down the road, in gaming, education, and mental health care.
How to Think Computationally about AI, the Universe and Everything by Stephen Wolfram. This TED talk is about how AI and the computational engine Wolfram Alpha change our understanding of the world. And how, when we think about things through the lens of computation, interesting concepts start to emerge. Stephen Wolfram likes to think of everything as computation and of us as explorers of a vast computational possibility space–a computational object he calls the "ruliad". And he argues that AIs will help us explore this ruliad and uncover the secrets of the universe. The links going out from this to various resources alone make this transcript worth a read.
Will Supercomputers and AI Drive Biotech Breakthroughs? by Maxx Chatsko. Biotech is producing a lot of data, and when you want to gain insights from that data you need a lot of computational power. But companies like Gingko Bioworks already know that and hence build their own hardware and server racks that are up for the challenge. The dream: A machine that you can feed properties of the protein you want, that then spits out the correct protein structure that would work as you envisioned. Something like ChatGPT, but for protein functions, not just structures like AlphaFold already gives us. Such a tool would allow to generate incredible new medicines and compounds. And it would be a huge step forward for the field of biotech.
🌌 Traveling 🌌
🎶 Song 🎶
Tank! by Seatbelts
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