Chapter
The Search for Good Predicates in Machine Learning.
44:21 - 51:37 (07:15)
Theore claims that if you use all available functions from Hilda's space, you won't need training data. However, if you only use a few good predicates, you'll need some training data. Thus, the search for good predicates remains an ongoing challenge in machine learning.
Clips
Claude Levistros discusses the existence of units in paradigms in theoretical literature and questions whether they would hold the same power without human interpretation.
44:21 - 45:34 (01:12)
Summary
Claude Levistros discusses the existence of units in paradigms in theoretical literature and questions whether they would hold the same power without human interpretation.
ChapterThe Search for Good Predicates in Machine Learning.
EpisodeVladimir Vapnik: Predicates, Invariants, and the Essence of Intelligence
PodcastLex Fridman Podcast
Theore proposes that if all possible predicates in Hilda's space are used, there is no need for training data.
45:34 - 47:52 (02:18)
Summary
Theore proposes that if all possible predicates in Hilda's space are used, there is no need for training data. To achieve this, a challenge was proposed to obtain state-of-the-art MNIST error rates using very few examples per digit.
ChapterThe Search for Good Predicates in Machine Learning.
EpisodeVladimir Vapnik: Predicates, Invariants, and the Essence of Intelligence
PodcastLex Fridman Podcast
This podcast episode discusses the possibility of an automated reasoning system for finding predicate functions and suggests that finding situations where existing theories cannot explain could be a starting point for discovering new predicates.
47:52 - 49:56 (02:03)
Summary
This podcast episode discusses the possibility of an automated reasoning system for finding predicate functions and suggests that finding situations where existing theories cannot explain could be a starting point for discovering new predicates.
ChapterThe Search for Good Predicates in Machine Learning.
EpisodeVladimir Vapnik: Predicates, Invariants, and the Essence of Intelligence
PodcastLex Fridman Podcast
The process of discovering contradictions and removing them is important to finding good predicates in AI, however, brute force is not the best way.
49:56 - 51:37 (01:41)
Summary
The process of discovering contradictions and removing them is important to finding good predicates in AI, however, brute force is not the best way. Logic reasoning is not enough for AI development as it requires more than just deduction for human-like learning.