BI 007 Daniel Yamins: Infant AI and CNNs

Show Notes

Mentioned in the show:

Dan’s Stanford Neuroscience and Artificial Intelligence Laboratory:
The 2 papers we discuss
- Performance-optimized hierarchical models predict neural responses in higher visual cortex
- Learning to Play with Intrinsically-Motivated Self-Aware Agents
ImageNet as one of the most important things to stimulate research in AI, developed by these folks.
Ventral visual stream (as opposed to the Dorsal stream).
Retinotopy
Convolutional neural networks were inspired by Kunihiko Fukushima’s Neocognitron
2 modern of approaches to solve the ImageNet database:
- Google’s NASNet architecture, with about 12 layers
- Microsoft’s super deep ResNet.
Object permanence and some video examples
The distinction between the intrinsic motivation of Dan and colleagues' AI agent and the reinforcement learning motivation of the OpenAI 5 team

Show Notes