Before It Worked
Curated by Claude · 11 videos · 6 hr 19 min
Most AI videos on HN are about what works now. This playlist goes back to the talks HN kept linking while neural nets were still a fringe bet: the debates, the demos and the pioneers explaining ideas almost nobody used yet. It runs in chronological order, from 1973 to 2017. The last two talks are critiques from people inside the field.
Play all on YouTube- 11:23:12
The Lighthill debate on Artificial Intelligence: "The general purpose robot is a mirage"
15K views • 8 years agoPierre-Yves OudeyerThe full 1973 BBC Lighthill debate on artificial intelligence, featuring James Lighthill, Donald Michie, Richard Gregory and John McCarthy. It documents the argument that preceded the 1970s AI winter and includes early robot demos.The full 1973 BBC debate in which Lighthill argues against McCarthy and Michie, the argument that came before the first AI winter, with vintage footage of the Freddy robot.
- 218:27
Richard Feynman: Can Machines Think?
2.5M views • 6 years agoLex ClipsAn excerpt of Richard Feynman's 1985 lecture Q&A about whether machines can think. He discusses the difficulty of defining what computers can never do and compares machine intelligence to machines that don't copy animals.In 1985 Feynman notes how hard it is to name anything a computer will never do, and HN still quotes that point in AGI arguments.
- 34:10
Karl Sims - Evolved Virtual Creatures, Evolution Simulation, 1994
797K views • 17 years agoMediaArtTubeKarl Sims's 1994 demonstration of evolved virtual creatures: block-bodied creatures whose bodies and controllers are evolved through selection, mutation and recombination to swim, walk and compete.Four minutes of evolved block creatures from 1994, which HN cites as the classic example of optimization exploiting simulator bugs.
- 452:52
Google Developers Day US - Theorizing from Data
27K views • 19 years agoGoogleDeveloperDayPeter 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.Norvig's 2007 case that more data beats cleverer models, the statistical approach that came before deep learning took over.
- 5
48:10Bay Area Vision Meeting: Unsupervised Feature Learning and Deep Learning
123K views • 15 years agoGoogle TechTalksAndrew 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.Andrew Ng in 2011 showing networks learn their own features from unlabeled data, a talk nine HN commenters recommended over two years as the simple overview of the coming wave.
- 6
4:12PREDATOR - Tracker that learns and improves
893K views • 15 years agoZdenek KalalA demo of PREDATOR (TLD), a real-time object tracker that learns the appearance of its target and improves while it tracks. The technology was later developed commercially.A 2011 vision demo that learns as it tracks, which HN remembers as primitive but effective and a reminder that pre-deep-learning methods could already impress.
- 7
1:00:19Brains, Sex, and Machine Learning
152K views • 14 years agoGoogle TechTalksGeoffrey 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.Hinton in 2012 explains dropout through sexual reproduction and spiking neurons, which HN calls a simple, elegant account of the technique.
- 8
9:05Speech Recognition Breakthrough for the Spoken, Translated Word
1M views • 14 years agoMicrosoft ResearchMicrosoft Research's Rick Rashid demonstrates a speech recognition and translation system, built on deep neural networks, that turns his spoken English into Chinese speech in 2012.Microsoft's 2012 live demo turning English speech into Chinese in the speaker's own voice, one of the first public signs that deep nets had arrived.
- 9
3:54Deep Visualization Toolbox
270K views • 11 years agoJason YosinskiA short demo of the Deep Visualization Toolbox, which shows the activations inside a convolutional neural network as it processes live camera input. Viewers see individual units responding to edges, faces and text.A 2015 look inside a convnet watching a live camera, with individual units lighting up for faces and text.
- 10
1:11:36Geoffrey Hinton talk "What is wrong with convolutional neural nets ?"
200K views • 9 years agotrwappersGeoffrey Hinton gives an MIT talk on what is wrong with convolutional neural networks and introduces his capsules project. The description points to the capsule network papers.Once convnets had won, Hinton came back to argue that pooling is a mistake and to pitch capsules instead.
- 11
23:20Ali Rahimi's talk at NIPS(NIPS 2017 Test-of-time award presentation)
78K views • 8 years agoObservantVidsAli 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.The ending: the field's most-quoted critique of itself, which argues that machine learning had become alchemy just as it started to work.