deep learning

Max Tegmark's MIT CBMM talk on why deep learning works so well. He argues that the laws of physics and the symmetries in physical data make it compressible into small neural networks, tying physics to network structure.

Andrew Ng presents unsupervised feature learning and deep learning, showing how algorithms can learn representations from unlabeled data instead of hand-designed features. He gives examples in vision and audio, including visualizations of features learned by successive layers.

Geoff Hinton's 2010 Google Tech Talk on deep belief networks, learning layers of representation one at a time from unlabeled data, and extending them with multiplicative interactions.

Valve's John McDonald presents at GDC 2018 on how Valve uses deep learning (VACnet) to detect cheaters in Counter-Strike: Global Offensive. He covers training on human-reviewed replays and how model output is confirmed by manual review.

A GDC 2018 talk by Valve's John McDonald on VACNet, a deep learning system that analyzes CS:GO gameplay to detect cheaters. It covers how the model is trained on human-reviewed Overwatch cases and how its output feeds human moderation.

Bill Dally of NVIDIA gives a Hot Chips 2023 keynote on the challenges of building hardware for deep learning. He traces how GPUs have been adapted for machine learning, especially training.

Ali Rahimi accepts the NIPS 2017 Test of Time Award and argues that machine learning has become alchemy, with practice running ahead of understanding.

Ali Rahimi's NIPS 2017 Test-of-Time award talk, in which he argues that modern machine learning has become alchemy, working without being rigorously understood. He calls for more rigor and understanding in the field.

Geoffrey Hinton's Google Tech Talk linking machine learning to biology, explaining dropout as a form of model averaging and why sexual reproduction and stochastic spiking neurons may prevent overfitting.

Chris Lattner and Brennan Saeta present Swift for TensorFlow at the TensorFlow Dev Summit 2019, covering language-integrated automatic differentiation and tooling for machine learning in Swift.

Simon Knowles, co-founder and CTO of Graphcore, talks about "intelligence processors", chips designed specifically for machine learning. The talk was given at the RAAIS 2017 summit and covers Graphcore's architecture.

UC Berkeley EECS colloquium by OpenAI's John Schulman on reinforcement learning from human feedback. He covers why RL helps with truthfulness, hallucination, and the limits of behavior cloning.

Demis Hassabis of DeepMind speaks at Google Zeitgeist, saying human-level AI is decades away but the debate should start now. He covers DeepMind's focus on game-playing agents.

James Mickens's humorous USENIX Security keynote argues that improving security and trusting opaque technology like machine learning is far harder than the industry admits.

Veritasium's Derek Muller gives a Perimeter Institute livestream on how AI might change teaching and learning. He draws on past technology shifts and cognitive science to argue that learning requires effortful repetition.

Jose Valim introduces Nx, the project bringing numerical computing to Elixir, covering multi-dimensional tensors, automatic differentiation and JIT compilation to CPU and GPU.

Andrej Karpathy, then Tesla's head of AI, presents at ScaledML 2020 on how Tesla builds full self-driving using vision-based neural nets and fleet data collection. He covers training custom detectors and neural nets gradually replacing traditional code.

Chris Lonsdale presents a TEDx talk on how adults can learn a language in about six months, centering on immersion, comprehensible input and a practical learning methodology.

George Hotz of Comma AI gives a 2017 conference talk on lessons from building a self-driving system. He argues about how driving should be defined and why the common approach to autonomous vehicles is flawed.

Peter Norvig's 2007 Google talk on theorizing from data, covering what huge datasets can do for language understanding, translation, information extraction and inference. It argues that more data often beats cleverer models.

Peter Norvig, then Director of Research at Google, talks at Startup School 2008 about leveraging large amounts of data and about technical challenges in Google projects such as image search. A central idea is that data can beat hand-written programming.

Chris Stucchio presents the ethical principles behind algorithmic decision making and shows with examples from lending, criminology and education that they cannot all be satisfied at once. The talk frames AI fairness as a set of quantitative tradeoffs.

Mobileye co-founder and CTO Amnon Shashua explains what goes into sensing for autonomous driving and whether end-to-end deep learning can work for it. He argues for decomposing the problem.

A Google Tech Talk by Laurens van der Maaten, co-creator of t-SNE, on visualizing high-dimensional data by mapping it to two dimensions. It explains how the technique works and how it compares with other visualization approaches.

A DeepSec 2010 conference talk by Ralf Philipp Weinmann on attacking the baseband processor of mobile phones rather than the application processor. It covers how open-source GSM base stations make malicious-tower attacks practical against neglected GSM stacks.

Andrej Karpathy's Databricks talk on Software 2.0, the shift from human-written code to code produced by neural network training. It covers how developer work moves toward curating datasets, the advantages and challenges, and opportunities for new tooling.

Ilya Sutskever gives a Simons Institute talk on why neural networks generalize, framing learning as finding short descriptions, like compression. It was part of the Large Language Models and Transformers program.

Rich Hickey's philosophical talk from the first Clojure Conj in 2010. He argues that good design comes from deliberate, away-from-keyboard thinking, and describes how to feed the background mind so it solves problems.

Mike Rowe, host of Dirty Jobs, shares stories from messy real-world jobs in this TED talk. He argues that hard work has been unjustly degraded in society.

Rick Houlihan's AWS re:Invent 2018 deep dive on advanced DynamoDB design patterns. It covers index overloading, GSI sharding, composite keys, graph processing and transactions.

Sal Khan of Khan Academy argues that schools push students forward with gaps in their foundations, and that students should instead master concepts at their own pace. He compares it to building a house on an unfinished foundation.

Jim Keller's talk at the 2020 Scaled Machine Learning conference on Moore's Law in the age of AI chips. He compares scalar, vector and GPU-style parallel approaches to computation.

Dragan Djuric presents Clojure libraries for GPU and numerical computing, including ClojureCL, ClojureCUDA and Neanderthal. He shows how a dynamic REPL workflow can be combined with low-level hardware acceleration for machine learning and data science.

Josh Kaufman's TEDx talk argues that you can reach practical competence in almost any skill with about 20 hours of focused practice, rather than the 10,000 hours often quoted. He describes how becoming a parent led him to rethink how adults learn, and lays out a short method for rapid skill acquisition.

Eduardo Briceno gives a TEDx talk on mindset, contrasting fixed and growth views of ability. He argues that believing abilities can be developed encourages challenge, deliberate practice and grit.

John Ousterhout, Stanford professor and creator of Tcl, presents the ideas of his book A Philosophy of Software Design in a Talks at Google session. He covers managing complexity, deep modules and interfaces, and what he teaches in his Stanford software design course.

Brian Karis's SIGGRAPH 2021 talk explains how Unreal Engine 5's Nanite virtualized geometry works. It covers mesh import, the data structures, streaming, decompression, culling, software and hardware rasterization, and shading.

Andrej Karpathy gives a one-hour general-audience introduction to large language models, covering what they are, how they are trained, comparisons to operating systems, and security challenges such as jailbreaks.

Conal Elliott presents his paper on automatic differentiation, deriving a simple, generalized AD algorithm from a natural specification. Varying the representation of derivatives yields forward and reverse mode algorithms that avoid the complexity and statefulness of backpropagation.

Andrej Karpathy walks through how GPT assistants are trained, from tokenization and pretraining to supervised finetuning and RLHF, and gives practical prompting strategies and a view of the tooling ecosystem.