1 / 22

Near- and Far-Term AI

Near- and Far-Term AI. Dr. Cynthia Matuszek cmat@umbc.edu University of Maryland, Baltimore County http:// iral.cs.umbc.edu. About Myself. Assistant Professor at UMBC Graduated from University of Washington 2014 Robotics and natural language processing “Fair” technology

nay
Download Presentation

Near- and Far-Term AI

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Near- and Far-Term AI Dr. Cynthia Matuszek cmat@umbc.edu University of Maryland, Baltimore County http://iral.cs.umbc.edu

  2. About Myself • Assistant Professor at UMBC • Graduated from University of Washington 2014 • Robotics and natural language processing • “Fair” technology • E.g., the effect of Google search on gender stereotypes • General topic: Artificial Intelligence, and what we can do with it

  3. This Talk A little about AI My work AI in education (now and soon) Useful “semi-AI” technologies Why AI, why technology, why now

  4. So… What is AI?

  5. Artificial Intelligence • Strong AI: mental/thought capabilitiesequal to (or better than) human • Weak (bounded) AI: intelligent actions or reasoning in some limited situations • “Human-level” intelligence • In what situation? • Internally? • Self-awareness / Consciousness • Autonomy

  6. “Intelligence” is Problematic • Intelligence, self-awareness, autonomy • How do we measure these? • Is there something ineffable missing? • What? • What’s an ‘intelligent action’? • In practice, ‘previously human only’

  7. A Better Definition • Artificial intelligence is the study of getting computers to do useful, complex things that make people’s lives easier or better • Tasks that are: • Repetitive • Dangerous • Under-resourced • …

  8. My Research • Artificial intelligence • How to get computers to behave in ways that we would consider to be “intelligent?” • Human-Robot Interaction (HRI) • How can we put robots in human spaces? • Robotics • How can we go from industrial robots to useful robots in human environments? (Schools, cars, homes…) • Natural Language Processing • How can computers learn to understand and speak human languages (English)?

  9. Background: Robots Now • Robots now: • Expensive • Complex • Special-purpose • Environments • Dedicated • Constrained • Use and Management • Controlled by trained experts • Slow and expensive to reconfigure/repurpose

  10. Home Robotics Now • As technology improves: • Smaller, cheaper, more capable • But still: • Very special purpose • Difficult (or impossible)to repurpose/modify

  11. Natural Interactions “That’s my yellow teapot!” • Natural language based HRI • Natural, intuitive, unscripted • Broadly useful • Already widely known • Learning language for specific tasksand users

  12. Language ↔ Robotics “apple” • How can language be grounded in physical, real-world concepts? • Grounded Language Acquisition: • Learning language from interaction with world • Learning about the world from language • Robotics  NLP • NLP gains from having data source (learn “green” without it!) • Robotics gains because language is a great interface!

  13. Learning About the World This is a red thing that you can eat, but don’t eat these blue ones red = blue = (eat??) • Learning novel language describing concepts • Formal language fully defines world • turn-left, red(x), … • World modeling: new language and conceptsfrominteractions • Associated with language • Grounded in percepts

  14. Targeted Applications “Could you please...”

  15. AI in Education

  16. AI at UMBC • Machine Learning • Robotics • Natural Language Processing • Vision and Visualization • Knowledge Representation • Planning and Inference • AI and Security • Fair and Explainable AI

  17. Current AI: TAs • AI TAs • Georgia Tech: Jill Watson • “One of the main reasons many students drop out [of online courses] is because they don’t receive enough teaching support. We created Jill as a way to provide faster answers and feedback.” • 97% accuracy answering questions • When confidence is high, answers students directly • Students didn’t suspect

  18. Automated Grading • Bubble tests, multiple choice, T/F • Short answer: harder, but pretty effective • Essays? • State of the art: • Input (many) examples of good and bad essays • Machine learn a grading model • Model used to grade papers • 94% agreement with human graders • Can it be gamed?

  19. AI Tutoring Available, personalized, broad Progressive, responsive difficulty Example: Duolingo language learning Estimates student’s current state Chooses sentences that will help learn Chatbot- or interface-based

  20. Useful Now Gmail! Piazza Catme Blackboard(…)

  21. Useful Now Plagiarism detection Hybrid/online classes Telepresence Skype

  22. Why? • Education has repetitive tasks • Never enough teachers (or TAs, or tutors) • Student-specific support • Good tools make us better faculty • More available • More informed • More “human” • Modeling tech use for students!

More Related