I want to be stronger. This is true in general, but in this case specifically I want to save the world and so need the resources to do so. The current main candidate for X-risk seems to be AI, so it would be rational to power up in order to help in that epic battle. So this is a list of skills that would be useful, along with sources from which to learn more. Sort of a skill tree?

Want to learn

AI Safety

• AGI Safety Fundamentals 2

Calculus

• Optimization problems. Pity I didn’t listen during the simulation classes in university…
• Multivariate differentials

Deep learning

• fast.ai book

Entropy

• Entropy and the Time Evolution of Macroscopic Systems

Game theory

• Games and Information

Linear Algebra

• Linear Algebra Done Right

Probability

• Probability Theory: The Logic of Science - this explains why Bayes was right, starting from a few core assumptions, all the way to showing how to do science properly

Probabalistic programming

• PyMC? Pyro?

Proofs

• What would be a good place to learn how to do mathematical prooving?
• TLA?

Reinforcement learning

• Maybe go through these?

Transformers

• Find a good explanation for them

Writing skilz

• Probably mainly a matter of practice?

Want to make

AI Safety

• A good god - obviously
• A dependency graph of what is required for alignment?
• A general overview of what won’t work and why
• Reviews/summaries of AI/alignment papers

Deep learning

• Anime recommendation system - input your ratings for various anime, get a list of predicted ratings
• Edible plant recognizer - input a picture of a plant, get a predicted species name
• bird song / bat sound recognizer - input a recording of a flying critter, get the species name out
• Working neural nets from scratch - various architectures to get a good grasp on them

Game theory

• Run a load of MCMC experiments to really grok competition
• The Darwin game

Reinforcement learning

• Blackjack player - David Silver’s assignment from the RL lectures

Probabilistic programming

• Analyze what effects animal mortality on roads
• Simulate how resinging effects song rates, i.e. rather than manually simulating agents, do it via e.g. Pyro

Transformers

• Implement GTP2?

Writing

• More blog posts
• Paper summaries

Learned

AI safety

• AGI Safety Fundamentals

Ecology

• A bachelor degree in environmental biology should suffice for the basics

Calculus

• University over 10 years ago - the basics are there, but only enough to know where to look for more
• Better explained - gives an intuitive understanding of the basic ideas. Doesn’t go into the nitty gritty of how to solve derivatives, but what it does explain is enough to actually understand what’s going on. It also helps that it makes the quotient rule an obvious and easily derived equation

Game theory

• The Strategy of Conflict

Linear Algebra

• University over 10 years ago - the basics are there, but only enough to know where to look for more

Programming

• Python
• Clojure
• A masters degree (sans the actual degree…) and 10 years of experience

Neural networks

• CS231n - a Stanford class for NNs - a very good primer to introduce the basic ideas
• Neural Networks and Deep Learning - a whole book going into NN details from scratch

Rationality

• The Scout Mindset
• Thinking, Fast and Slow
• The Sequences

Reinforcement learning

• David Silver - Reinforcement Learning - a basic overview of what it’s all about