Rethinking software engineering in the AI era
summary
Argues that conventional rule‑based feature engineering gives way to learned models, replacing hand‑crafted descriptors with convolutional neural nets trained on massive data and compute, inspired by layered visual cortex processing. Cites ImageNet benchmarks where pre‑deep methods hit ImageNet top‑5 ~26% and then AlexNet 2012 breakthrough produces a large accuracy jump. Emphasizes using scaled datasets, GPUs and end‑to‑end learned feature hierarchies to get practical vision systems.
AI-written summary · about
related talks
27mPattern Analysis: Learned vs. Engineered
01h 00mDeep Learning Revolution: Rethinking Machine Learning Pipelines
04mInterview
29mDeep Learning, Neuroscience and the future of AI
41mDeep Learning to Solve Challenging Problems
04mUsing Computer Vision to Understand Big Visual Data
56mTutorial on Deep Learning II
40mArtificial Intelligence in Computer Vision