Josh Tenenbaum: Engineering & reverse-engineering human common sense
Distinguished Lecture Series: Josh Tenenbaum
Title: Engineering and reverse-engineering human common sense
Abstract:
If our goal is to build knowledge representations sufficient for human-level AI, it is natural to draw inspiration from how knowledge works in human minds. I will talk about insights AI researchers might draw from recent progress by cognitive scientists studying common sense knowledge and reasoning. First, although many AI researchers have approached common sense through natural language, we should consider starting (as humans do) with a core of non-verbal common sense, already present in infants 12 months of age and younger before they learn to produce or comprehend any sophisticated language. This is knowledge about physical objects, intentional agents, and their causal interactions -- an intuitive physics with concepts analogous to force, mass and the dynamics of motion, and an intuitive psychology with concepts analogous to beliefs, desires and intentions, or probabilistic expectations, utilities and plans -- which is grounded in perception but abstracts conceptually well beyond our direct perceptual experience. Second, we should explore approaches for language understanding, language-based reasoning, and learning that build naturally on top of this common-sense core. These considerations favor an approach to knowledge representation and reasoning based on probabilistic programs, a framework that combines strengths of traditional probabilistic and symbolic approaches. Building tools for language understanding and learning on top of probabilistic programs poses major engineering challenges, which I cannot claim to have much insight into but which I hope we might discuss together following the talk.
Видео Josh Tenenbaum: Engineering & reverse-engineering human common sense канала Allen Institute for AI
Title: Engineering and reverse-engineering human common sense
Abstract:
If our goal is to build knowledge representations sufficient for human-level AI, it is natural to draw inspiration from how knowledge works in human minds. I will talk about insights AI researchers might draw from recent progress by cognitive scientists studying common sense knowledge and reasoning. First, although many AI researchers have approached common sense through natural language, we should consider starting (as humans do) with a core of non-verbal common sense, already present in infants 12 months of age and younger before they learn to produce or comprehend any sophisticated language. This is knowledge about physical objects, intentional agents, and their causal interactions -- an intuitive physics with concepts analogous to force, mass and the dynamics of motion, and an intuitive psychology with concepts analogous to beliefs, desires and intentions, or probabilistic expectations, utilities and plans -- which is grounded in perception but abstracts conceptually well beyond our direct perceptual experience. Second, we should explore approaches for language understanding, language-based reasoning, and learning that build naturally on top of this common-sense core. These considerations favor an approach to knowledge representation and reasoning based on probabilistic programs, a framework that combines strengths of traditional probabilistic and symbolic approaches. Building tools for language understanding and learning on top of probabilistic programs poses major engineering challenges, which I cannot claim to have much insight into but which I hope we might discuss together following the talk.
Видео Josh Tenenbaum: Engineering & reverse-engineering human common sense канала Allen Institute for AI
Показать
Комментарии отсутствуют
Информация о видео
Другие видео канала
Transformers as Soft Reasoners over Language | AI2Why AI2?OpenBot: Turning Smartphones into Robots | Embodied AI Lecture Series at AI2MIT AGI: Building machines that see, learn, and think like people (Josh Tenenbaum)Beyond the Label: Robots that Reason about Object Semantics | Embodied AI Lecture Series at AI2The incredible inventions of intuitive AI | Maurice ContiComputers with common sense | Doug Lenat | TEDxYouth@AustinNeuroscience, AI and the Future of Education | Scott Bolland | TEDxSouthBankEEVblog #684 - Ness SMT Manufacturing & Assembly Factory TourConsciousness in Artificial Intelligence | John Searle | Talks at GoogleBringing AI and machine learning innovations to healthcare (Google I/O '18)Artificial Intelligence, the History and Future - with Chris BishopGary Marcus - Artificial General Intelligence: Why Aren't We There Yet?Josh Tenenbaum - The cognitive science perspective: Reverse-engineering the mind (CCN 2017)Creating Human-level AI: How and When?What are the differences between neural networks and the brain?Ian Goodfellow: Generative Adversarial Networks (GANs) | Lex Fridman Podcast #19MIT Quest for Intelligence Launch: Scaling AI the Human WayArtificial Intelligence Colloquium: KeynoteMIT AGI: Cognitive Architecture (Nate Derbinsky)