In one sentence: Danielle's husband Nuno, an economist and management consultant, explains how he taught himself machine learning in evenings and weekends, and the two of them discover that the "process vs result" split they usually draw in investing is blurrier than it looks.
Key ideas
- Learning is an investment of scarce free time. Nuno chose machine learning for two reasons: a natural affinity (his economics and statistics background) and a business reason (he could serve clients better). Pick what you will keep at because it pays off for you. [01:05–03:05]
- A three-step self-study plan. (1) Understand the types of algorithms and where each works. (2) Learn to code them (he used R, an open-source statistics language). (3) Run an R&D phase on real data to show you can produce useful insight. [04:00–08:00]
- Learning is never finished. There is always another algorithm or project, even for experts. Danielle agrees: investing practice has no finish line either. [08:02–09:00]
- Online forums do much of the heavy lifting on the hard parts. For code, he searched, found a worked answer and adapted it. Reuse what others have solved and keep your effort for the part that is yours. [05:00–07:00]
- The painful part is the means, not the end. He did not enjoy coding or dense academic papers, but kept going because he wanted the answers they unlock. [10:00–14:00]
- Results-driven vs process-driven. Nuno is driven by the insight; Danielle by the daily chasing-down of ideas. Settling the "semantics": the process is full of small results (each interesting finding), and the big result is a bonus. [14:03–19:00]
- Repetitive steps are the least fun part, but learning something new inside the process makes it enjoyable. [17:02]
How it maps to RuleOne
- This is a note about the practice behind Rule #1 rather than a method. The site's screen and stock pages exist to take over the mechanical parts (collecting numbers) so your time goes to the part that needs judgement.
- It sets up 424, on AI as a research tool.
Buffett, Munger and Graham links
- Munger's habit of constant learning, described in Poor Charlie's Almanack (the "learning machine" idea), matches the "never completed" point. This is my link, not something said in the episode.
Words to know
- Algorithm: a repeatable set of steps for turning data into a prediction or decision.
- R: a free programming language for statistics.
Try this
Write down the three steps you would take to learn one new investing skill (for example reading a balance sheet). Do step 1 for 30 minutes this week with a company from All stocks.
Check yourself
- What were Nuno's three steps in learning AI?
Answer
Understand the algorithms, learn to code them, then do R&D on real data. - How did Nuno and Danielle reconcile process and result?
Answer
The process is made of many small results (insights), so the two are not as separate as they seem. You still need to like the process well enough.
Short quotes
"There's so much investing practice to be had from conversations about things totally separate from investing practice." (Danielle, ~18:50, auto-transcribed)