Is Word Learning Innate or Learned? Recent Evidence Suggests It Might be the Latter (with Wai Keen Vong)

Unless you live under a rock, you've no doubt heard about that recent Science paper showing that an artificial neural network trained on visual and linguistic data from a child's egocentric recordings was able to map words with their referents. What makes this paper particularly significant is that it demonstrates that there is sufficient structure in the environment to support children's ability to learn words, without the need for strong inductive and innate biases or domain-specific knowledge.

Well, guess what? In this episode, we were joined by the lead author of that paper, Dr. Wai Keen Vong! We cover lots of ground in this episode — we talk about the paper, its motivations, Wei Keen's journey in science, how one might go about learning more about computational modeling, among other things.

Links

Vong, W. K., Wang, W., Orhan, A. E., & Lake, B. M (2024). Grounded language acquisition through the eyes and ears of a single child. Science. Link