in this newsletter: month update, some thoughts (on computer limits and evolving learning patterns), my reading list, something interesting, what’s next. feel free to jump around.
I spent the beginning of my month in Bogota for Devcon. I learned a lot about new ideas and applications people are excited about and made some incredible new friends too. Mostly working in research, I never reflected on the different ideas that people were actively working towards making mainstream, so talking to people at the event and listening to some of the talks was an exciting experience for me.
I also spoke at Sustainable Blockchain Summit, a Devcon side event, about collective action and incentivizing sustainability with the blockchain. I had an amazing climate roundtable with people leading the ReFi movement. There, we created a document to talk about timelines and problems that need to be solved.
I ended the month with the Gairdner Luncheon- I talked to Dr. Guy Rouleau, Director of the Montreal Neurological Institute-Hospital. We had a great conversation about Open Science - something he’s advocated for his entire career, the future of open sourcing progress, and implications of morality.
One of my favourite talks from Dr. Deborah Cook, discussed how many parts of science are hard to work on because people actively discourage the studying of it. Most of her work was on bringing research into clinical practice in the end of life stage. From her talk, I realized emotional and social implications of end of life care are just as important as leading with science to improve quality of life.
So what am I up to now? For the past 2 months, I was exploring reinforcement learning and how to leverage it to address real world challenges. Now, I’m working on an ML-enabled biomechanics tool so I’m learning to deal with data and automatically create insights from biological information. While I’m acquiring a more general skillset, I thought it would be fitting to work on using graph neural networks for drug discovery. I have a substack blog, which details more of how this works. In the next month, I’m planning to write more short form notes as I’m documenting my progression of knowledge in the space.
#1 - on the limit of computers
One of the most compelling arguments on the limits of computer intelligence was the idea that biological and physical things we experience today aren’t a product of intense computation, but instead, just happen.
An interesting Twitter thread I read a few days ago talks about how nature acts with complex mathematical models but it doesn’t require the intense computational power. The other question to ask is - Where does the universe get its computational power from?
From the thread: “It now seems conceptually possible to me that the brain produces intelligent behaviour without having to explicitly compute it, and that building machines to explicitly compute intelligent behaviour could be infeasible to reach AGI.”
From many episodes of the Lex Fridman podcast, I’ve seen AI research scientists talk about how the universe is a big computational puzzle. From beginning to read some of Nick Lane’s work and continuing to better understand the parallels between AI and biology, I’m excited to see how my opinions on this evolve.
#2 - evolving to learning patterns and quick decision making
On an episode of TKP, Reid Hoffman talks about principles of decision making and how to constantly evolve your learning pattern amongst other things.
He talks about how to scale a company.
“You have to be constantly evolving your learning pattern. What got you here, won’t get you there. Making a decision faster is always better.”
In situations, it’s important to ask: how much time and costs would it take to get more information to make a better decision?
Or more precisely, it’s better to ask how do you make decisions fast and well?
Be an explicit leaner: try to learn things in a way that you could explain them to others, this internal process will help you work towards decisions faster and easier
With every question, run a practice analysis on your immediate answer: who else would you talk to, what other data could you get, what are the consequences of the decision—learn your set of questions to run through
Identify irreversible consequences: the more irreversible the consequences, the higher degree of certainty you should seek (not all decisions can be made fast)
what i read this month & what i’m reading now
I thought this section would be good to include for people who may be reading similar things - maybe we could talk about some ideas from these pieces + would love to hear what’s been on your list recently.
until next month..
Here’s what I’ll be up to:
working on a generative model for structure-based drug design
participating in the open ai climate hackathon
exploring and detailing climate solutions on medium
on a mainstage panel at W3B 2022 with Don Tapscott (lmk if you’re going and we can meet there)
Thank you for reading. If you know someone who might enjoy this update send them this link: https://anyasingh.substack.com
Thanks for reading anya updates! Subscribe for free to receive new posts and support my work.