Augmenting Human Capabilities to New Dimensions
summary
Current deep learning needs huge labeled datasets and humans in the loop, fails at quick adaptation and object discovery, and lacks the slow deliberative planning humans use; AlphaGo shows planning works but relies on a human-coded model. New research claims to combine learned internal models with planning and merge neural and symbolic methods to discover objects from raw pixels. The promised result is an AI co-worker that learns quickly, can act autonomously and collaborate with other AIs — research exists but no product ships yet.
AI-written summary · about
chapters (5)
- 0:00Introduction and AI overview
- 1:18Current AI limitations and human role
- 5:19System 2 thinking and planning in AI
- 8:11Future AI co-workers and collaboration
- 11:29Conclusion and future vision
related talks
21mWhen the machines out-smarts us
04mFrom big data to human-level artificial intelligence
06mInterview
48mThe Perils and Promise of Artificial Intelligence
13mArtificial General Intelligence: Why aren't we there yet?
24mAI as a Teammate: Redefining Agile for Human–AI Collaboration
54mAI vs Humans: Rewriting the Rules of Success
32mAugmented Intelligence - A Marriage between Machine and Human