Human Learning
summary
Claims that IT knowledge has a short half-life of knowledge and that attention limits learning to 5 to 9 chunks drive the talk, backed by concrete models and numbers: a classroom memory demo (17 repetitions to fix 12 syllables), algorithms used (naive Bayes with Laplace smoothing, KNN, SVM, clustering, knowledge graphs), and training methods (deep learning and reinforcement learning with human feedback as used for ChatGPT). Operational costs and tradeoffs appear too: training reported in the millions and runtime in the hundreds of thousands, linking biological memory models to practical representation and redundancy principles.
AI-written summary · about
chapters (10)
- 0:00Human learning today
- 3:14Human learning in tech
- 7:51Brain processing and storage
- 15:32Data representation principles
- 20:53Schemas in the brain
- 26:08Knowledge types and transfer
- 28:36Self-regulated learning
- 31:48Machine vs human learning
- 33:53Intelligence beyond IQ
- 38:48Cost and purpose of learning
related talks
26mLearn like a machine
37mActive Teaching, Human Learning
01h 07m(How) Are We Staying Smarter Than AI?
32mA New Kind of Learning Systems
28mPreference Learning from Minimal Human Feedback for Interactive
39mComputational Thinking on Learning Models
26mHow Machine Learning Can Help
41mArtificial Intelligence: Let's Think About Learning