I love Sean Carroll’s new Mindscape Podcast because its full of little gems like the conversation in the last four minutes of Episode 16: Coleen Murphy on Aging, Biology, and the Future.
For context, Coleen is a biologist who studies aging in the C. Elegans nematode. Anyway, the last topic was about how C. Elegans seems to be able to pass down learning genetically. Specifically, her lab trains a parent to avoid a normally fine food, which the offspring then also avoid without the same training. As she hypothesizes, that would be a neat trick if a new threat emerged that tasted1 like something that’s normally safe, but without being as heavy-handed as evolving an aversion to an historically safe food that might become so again.
This got me excited because it sounds a lot like Lamarckism, a theory on the “inheritance of acquired characteristics” proposed in the early 1800s. And if you’re wondering how I know about that at all,3 it’s something I learned about during the peer review of my first journal article on a Genetic Algorithm and Calculus of Variations-Based Trajectory Optimization Technique.
In that paper we coupled two optimization techniques. The higher-level optimizer was a classical genetic algorithm, but each genome was passed off to a classical, gradient-following optimizer before the objective function was evaluated. I ran it in two modes and compared the results. In the first, the genome was untouched,4 but in the second, the gradient optimizer was allowed to change the genome before breeding the next generation.
When I got my first round of peer review feedback, one of the reviewers said these sounded like Baldwinian and Lamarkian evoluationary strategies, and after reading some 200 year old biology papers, I realized they were right.5
The Lamarkian approach was vastly superior in my application, and I remember thinking it’s a bummer Lamarck was wrong about evolution, because I bet inhereted learning would be really useful to animals who hadn’t invented things like books.
Turns out™, he may have been onto something after all.