Chapter
The Limitations of Deep Learning in AI
Deep learning has been successful in certain areas, but it has not resulted in significant progress in natural language understanding or common sense reasoning. AI has yet to figure out how to perform the functions that human brains do effortlessly with very little energy consumption.
Clips
This episode discusses the accessibility of intuition for artificial intelligence systems and how it has not yet been encoded in machine interpretable form.
49:27 - 50:12 (00:44)
Summary
This episode discusses the accessibility of intuition for artificial intelligence systems and how it has not yet been encoded in machine interpretable form. The closest attempt was the CYC, psych system.
ChapterThe Limitations of Deep Learning in AI
EpisodeGary Marcus: Toward a Hybrid of Deep Learning and Symbolic AI
PodcastLex Fridman Podcast
There is no commercially valuable model from 2010 in deep learning, as they are all considered failures.
50:12 - 52:10 (01:58)
Summary
There is no commercially valuable model from 2010 in deep learning, as they are all considered failures. While there may be room for improvement, the widespread belief is that psychologically informed AI simply does not work.
ChapterThe Limitations of Deep Learning in AI
EpisodeGary Marcus: Toward a Hybrid of Deep Learning and Symbolic AI
PodcastLex Fridman Podcast
Deep learning has led to significant progress for perceptual classification problems, but it has not offered any real advancement in areas such as natural language understanding or common sense reasoning.
52:10 - 56:57 (04:47)
Summary
Deep learning has led to significant progress for perceptual classification problems, but it has not offered any real advancement in areas such as natural language understanding or common sense reasoning. While deep learning has been successful, it cannot replicate the complex processes of the human brain, which uses much less energy and manages to achieve a lot more than even the most advanced AI.