Francis Ho
sunfishstanford.bsky.social
Francis Ho
@sunfishstanford.bsky.social
Silicon Valley VC at Walden Catalyst Ventures. Invest in DeepTech and TechBio startups. Ex McKinsey, JP Morgan Partners, Samsung, Inphi. Student of physics and math.
Many types of generative AI use ideas related to maximum likelihood, expectation maximization (EM), variational inference, etc. And the math involved is very cool. So I did a deep dive, and summarized what I learned in the post below. 👇️

freeenergy.blog/2026/EMnotes...
A step-by-step explanation of how EM (expectation-maximization) is used for training a generative AI model
These notes are based on CMU Prof. Bhiksha Raj’s lecture video on generative models and variational autoencoders and his lecture slides, specifically the section about expectation maximization and var...
freeenergy.blog
January 6, 2026 at 6:26 AM
In RL (reinforcement learning), learning how to prove the Performance Difference Lemma helped me to better understand some of the basic concepts in policy gradients, Markov Decision Processes, sampling trajectories from a policy, etc.

freeenergy.blog/2026/RLPerfo...
The Performance Difference Lemma in Reinforcement Learning
Sharing ideas and learnings in math, science, and tech
freeenergy.blog
January 5, 2026 at 12:39 AM
In William Gibson's 1986 short stories collection Burning Chrome, from Bruce Sterling's preface:

…we see a future…drawn from the modern condition…not from the shopworn formula of robots, spaceships, and the modern miracle of atomic energy, but from cybernetics, biotech, and the communications web…
December 28, 2025 at 5:50 PM
One way to interpret the explosion in demand from AI datacenters for electricity : many businesses believe that they have ways to convert electricity to new services with large profits. #AI #energy
September 13, 2025 at 6:17 PM