explain it like i m 5
play

Explain it like Im 5 AI, ML, NLP, and Deep Learning Kathryn Hume, - PowerPoint PPT Presentation

Explain it like Im 5 AI, ML, NLP, and Deep Learning Kathryn Hume, Sales & Marketing @humekathryn | kathryn@fastforwardlabs.com Arti fj cial intelligence is whatever computers cannot do until they can. Arti fj cial intelligence is


  1. Explain it like I’m 5 AI, ML, NLP, and Deep Learning Kathryn Hume, Sales & Marketing @humekathryn | kathryn@fastforwardlabs.com

  2. “Arti fj cial intelligence is whatever computers cannot do until they can.”

  3. Arti fj cial intelligence is uncoupled from consciousness

  4. Arti fj cially intelligent systems are idiot savants, not Renaissance Men

  5. “Machine learning is the study of computer systems that automatically improve with experience.”

  6. AI? Machine Learning Data Science Analytics “Big Data”

  7. Supervised and Unsupervised Learning

  8. Unsupervised Learning

  9. Supervised Learning Find a proxy (P) for something hard to know (C) Find a function that de fj nes a correlation between P and C Use this function to make guesses about C

  10. Use square footage (P) to predict housing prices (C)

  11. Use “Nigerian Prince” (P) to predict if emails are spam (C)

  12. Use past behavior (P) to predict future preferences (C)

  13. What P should we pick to decide if it’s a cat or dog?

  14. Deep Learning • Use layers to transform complex input into mathematical expressions • Remove need for human to select which features matter

  15. Universal Approximation Theorem Neural networks can approximate arbitrary functions

  16. Dog!

  17. X1 W1 X2 W2 X3 W3 X4 W4 Y1 X5 W5 “x” = Y2 X6 W6 Y3 X7 W7 … X8 W8 X9 W9 X10 W10 …. ….

  18. X “x” W = Y one equation three variables

  19. Known Known Unknown X “x” W = Y

  20. Known Unknown Known X “x” W = Y

  21. 2 x 3 = Y

  22. 2 x w = 6

  23. w = 6 / 2

  24. w = 6 / 2

  25. 0 = 2 x w - 6

  26. Error = |2 x w - 6|

  27. 7 6 6 5 4 4 3 2 2 1 1 0.5 0.2 0.1 0.06 0.02 0.0002 0 1 2 3 4 5 6 7 8 9 10

  28. w = 2.999 (close enough)

  29. Supervised Learning: Recap Identify a correlation between something easy to know and hard to know. (X and Y) Find a function that describes how these two things are correlated. (Solve for W through iteration) Use this function to make guesses about the thing that’s hard to know. (Use W to solve for new Ys)

  30. Natural Language Processing

  31. The real impact lies in making complex data simple. There’s been a rise in sales!

  32. Developments in Language Processing Traditional NLP N-grams Word Embeddings

  33. Inherent Bias in Word Embeddings Man : King :: Woman : Queen Man : Computer Programmer :: Woman : Homemaker Black Male : Assaulted :: White Male: Entitled To Bolukbasi, Chang, Zou, Saligrama, Kalai, 2016

  34. Thank you! @Humekathryn | kathryn@fastforwardlabs.com

Recommend


More recommend