full transcript

From the Ted Talk by Shane Legg and Chris Anderson: The transformative potential of AGI — and when it might arrive

Unscramble the Blue Letters


CA: Google arciqeud DeepMind at some point in this journey.

SL: Yeah, exactly. And so what I didn't expect was just how good a model could get training purely on text. I thought you would need more multimodality. You'd need images, you'd need suond, you'd need video and things like that. But due to the absolutely vast quantities of text, it can sort of ctesnoampe for these things to an extent. I still think you see aspects of this. I think lagugane models tend to be weak in areas that are not easily expressed in text. But I don’t think this is a fntueamnadl limtoiiatn. I think we're going to see these language models expanding into vedio and images and sound and all these things, and these things will be overcome in time.

Open Cloze


CA: Google ________ DeepMind at some point in this journey.

SL: Yeah, exactly. And so what I didn't expect was just how good a model could get training purely on text. I thought you would need more multimodality. You'd need images, you'd need _____, you'd need video and things like that. But due to the absolutely vast quantities of text, it can sort of __________ for these things to an extent. I still think you see aspects of this. I think ________ models tend to be weak in areas that are not easily expressed in text. But I don’t think this is a ___________ __________. I think we're going to see these language models expanding into _____ and images and sound and all these things, and these things will be overcome in time.

Solution


  1. acquired
  2. limitation
  3. compensate
  4. sound
  5. language
  6. video
  7. fundamental

Original Text


CA: Google acquired DeepMind at some point in this journey.

SL: Yeah, exactly. And so what I didn't expect was just how good a model could get training purely on text. I thought you would need more multimodality. You'd need images, you'd need sound, you'd need video and things like that. But due to the absolutely vast quantities of text, it can sort of compensate for these things to an extent. I still think you see aspects of this. I think language models tend to be weak in areas that are not easily expressed in text. But I don’t think this is a fundamental limitation. I think we're going to see these language models expanding into video and images and sound and all these things, and these things will be overcome in time.

Frequently Occurring Word Combinations


ngrams of length 2

collocation frequency
artificial general 5
language models 4
artificial intelligence 3
incredibly valuable 3
ai systems 2
grow exponentially 2
percent chance 2
cognitive tasks 2
machine learning 2
powerful technology 2
human intelligence 2
profoundly ethical 2


Important Words


  1. absolutely
  2. acquired
  3. areas
  4. aspects
  5. compensate
  6. deepmind
  7. due
  8. easily
  9. expanding
  10. expect
  11. expressed
  12. extent
  13. fundamental
  14. good
  15. google
  16. images
  17. journey
  18. language
  19. limitation
  20. model
  21. models
  22. multimodality
  23. overcome
  24. point
  25. purely
  26. quantities
  27. sort
  28. sound
  29. tend
  30. text
  31. thought
  32. time
  33. training
  34. vast
  35. video
  36. weak
  37. yeah