Strong views, occasionally held

  • Increasing diversity in technology must be our first priority.
    • We can begin this by changing where venture capital investment goes. 
  • Our healthcare systems are failing patients directly.
  • AI-first drug discovery efforts will fail to produce useful results. Human drug developers have important intuition that cannot be replaced with current tech.
    • AI is ill-suited for novel discovery. Additionally our current understanding pharmacokinetics is non-determinist, compounding the problem. 
    • Brute force solutions are computationally intractable, the search space is simply too vast. 
  • The way we live is shaped by our education, so it’s important to get it right. At the same time, we value commodity technical education over rigorous, classical, liberal arts education.  This is bad, and has lead to the technocratic nightmares we are currently experiencing.
  • It is important to consider the efficiency of one's work when choosing it. 
    • Sometimes, a few of our efforts to improve the world cause overwhelmingly more impact than others. Attempts to do good follow a heavily skewed distribution: where the top few projects save and improve many, many more lives than others. This follows a Pareto distribution
    • Effective Altruism is probably right, although the results of unbounded consequentialism at extremes of cost functions are definitely not. 
  • Technology's obsession with Philosophy is both not genuine, and not a good thing. 
    • This is the misappropriation of a misunderstood tradition of knowledge used to justify horrors.
    • Public intellectualism among VCs, however, is probably not the worst thing and possibly a practical necessity.
  • Weak type systems make programming more fun, productive, and assessable. 
    • The future will be built in Python, not Java.
    • Domain expertise is now the thing that limits rate of discovery/interesting results, not the ability to write software. Weakly typed languages allow for domain experts to thrive.