full transcript

From the Ted Talk by Daniela Rus: How AI will step off the screen and into the real world

Unscramble the Blue Letters


So here's the idea. Can we bliud AI using inspiration from the math of these neurons? We have developed, together with my collaborators and students, a new approach to AI we call “liquid networks.” And liquid networks rtsules in much more compact and explainable solutions than today's traditional AI slutoions. Let me show you.

This is our self-driving car. It's trained using a traditional AI solution, the kind you find in many applications tdaoy. This is the dhrsaoabd of the car. In the lower right coenrr, you'll see the map. In the upper left corner, the carmea input stream. And the big box in the middle with the bnikinlg lights is the decision-making engine. It consists of tens of thousands of artificial neurons, and it diedces how the car should steer. It is impossible to correlate the activity of these neurons with the behavior of the car. Moreover, if you look at the lower left side, you see where in the image this decision-making engnie looks to tell the car what to do. And you see how noisy it is. And this car drives by looking at the bsuehs and the trees on the side of the road. That's not how we dvire. People look at the road. Now contrast this with our liquid network solution, which consists of only 19 neurons rather than tens of tonhdsaus. And look at its attention map. It's so clean and focused on the road hoirzon and the side of the road. Because these moedls are so much smaller, we actually understand how they make decisions.

Open Cloze


So here's the idea. Can we _____ AI using inspiration from the math of these neurons? We have developed, together with my collaborators and students, a new approach to AI we call “liquid networks.” And liquid networks _______ in much more compact and explainable solutions than today's traditional AI _________. Let me show you.

This is our self-driving car. It's trained using a traditional AI solution, the kind you find in many applications _____. This is the _________ of the car. In the lower right ______, you'll see the map. In the upper left corner, the ______ input stream. And the big box in the middle with the ________ lights is the decision-making engine. It consists of tens of thousands of artificial neurons, and it _______ how the car should steer. It is impossible to correlate the activity of these neurons with the behavior of the car. Moreover, if you look at the lower left side, you see where in the image this decision-making ______ looks to tell the car what to do. And you see how noisy it is. And this car drives by looking at the ______ and the trees on the side of the road. That's not how we _____. People look at the road. Now contrast this with our liquid network solution, which consists of only 19 neurons rather than tens of _________. And look at its attention map. It's so clean and focused on the road _______ and the side of the road. Because these ______ are so much smaller, we actually understand how they make decisions.

Solution


  1. dashboard
  2. bushes
  3. thousands
  4. build
  5. decides
  6. results
  7. corner
  8. blinking
  9. today
  10. models
  11. horizon
  12. drive
  13. camera
  14. solutions
  15. engine

Original Text


So here's the idea. Can we build AI using inspiration from the math of these neurons? We have developed, together with my collaborators and students, a new approach to AI we call “liquid networks.” And liquid networks results in much more compact and explainable solutions than today's traditional AI solutions. Let me show you.

This is our self-driving car. It's trained using a traditional AI solution, the kind you find in many applications today. This is the dashboard of the car. In the lower right corner, you'll see the map. In the upper left corner, the camera input stream. And the big box in the middle with the blinking lights is the decision-making engine. It consists of tens of thousands of artificial neurons, and it decides how the car should steer. It is impossible to correlate the activity of these neurons with the behavior of the car. Moreover, if you look at the lower left side, you see where in the image this decision-making engine looks to tell the car what to do. And you see how noisy it is. And this car drives by looking at the bushes and the trees on the side of the road. That's not how we drive. People look at the road. Now contrast this with our liquid network solution, which consists of only 19 neurons rather than tens of thousands. And look at its attention map. It's so clean and focused on the road horizon and the side of the road. Because these models are so much smaller, we actually understand how they make decisions.

Frequently Occurring Word Combinations


ngrams of length 2

collocation frequency
physical intelligence 9
liquid networks 6
traditional ai 6
digital world 2
physical world 2
server farms 2
ai solutions 2
differential equations 2
teach robots 2


Important Words


  1. activity
  2. ai
  3. applications
  4. approach
  5. artificial
  6. attention
  7. behavior
  8. big
  9. blinking
  10. box
  11. build
  12. bushes
  13. call
  14. camera
  15. car
  16. clean
  17. collaborators
  18. compact
  19. consists
  20. contrast
  21. corner
  22. correlate
  23. dashboard
  24. decides
  25. decisions
  26. developed
  27. drive
  28. drives
  29. engine
  30. explainable
  31. find
  32. focused
  33. horizon
  34. idea
  35. image
  36. impossible
  37. input
  38. inspiration
  39. kind
  40. left
  41. lights
  42. liquid
  43. map
  44. math
  45. middle
  46. models
  47. network
  48. networks
  49. neurons
  50. noisy
  51. people
  52. results
  53. road
  54. show
  55. side
  56. smaller
  57. solution
  58. solutions
  59. steer
  60. stream
  61. students
  62. tens
  63. thousands
  64. today
  65. traditional
  66. trained
  67. trees
  68. understand
  69. upper