There is no such thing as Artificial Intelligence
summary
Claim: there is no single “artificial intelligence” but many narrow tools. Traces history from Pascaline (1642) through 1956 Dartmouth, expert systems, machine learning and deep learning. Gives numbers: 98% accuracy on cats with ~100,000 images, AlphaGo runs on ~2,000 machines (≈1500 CPUs, 300 GPUs, TPUs) consuming ≈440 kW vs a human ≈20 W, and failure cases: Tay shuts down after 16 hours due to biased training data and over-adaptive dialogue. Argues autonomy level five is effectively impossible; recommends thinking of AI as augmented intelligence — a toolbox of specific, energy- and data-limited tools.
AI-written summary · about
chapters (36)
- 0:00Intro
- 0:43No such thing as Artificial Intelligence
- 1:45Definition of Artificial Intelligence
- 5:20Natural Language
- 7:25Expert Systems
- 8:50Machine Learning
- 9:24Data
- 13:07Deep Learning
- 14:07Tensorflow
- 16:07Data is biased
- 17:07Microsoft chatbot
- 18:07Logical bug
- 19:07Databag
- 20:07switchboard
- 23:07explicable
- 24:07fractals
- 25:07Bruno Mars
- 26:07Fractal
- 27:07Autonomous Cars
- 29:07Level 4 Cars
- 31:07WeMo
- 32:07The Gem
- 33:07Stops
- 34:07Stop sign
- 35:07Autonomous car
- 36:07Siri
- 37:07Augmented Intelligence
- 38:07Tools
- 39:07Genesis
- 40:07What happened
- 41:07Generational AI
- 42:07Mars
- 43:07Cats
- 44:07AI based on mathematics
- 45:07Artificial or augmented intelligence
- 46:07generative AI
related talks
13mArtificial General Intelligence: Why aren't we there yet?
15mMore than 70 years of AI history in 15 minutes!
48mArtificial Intelligence. Our journey towards singularity
39mThe Pocket Guide to Artificial Intelligence
42mDawn of the planet of artificial intelligence
19mAI: The current state of play
46mAI vs Humans Rewriting the Rules of Success
52mArtificial, yes; Intelligent, no.