full transcript

From the Ted Talk by Brian S. Lowery and Kylan Gibbs: What makes us human in the age of AI? A psychologist and a technologist answer

Unscramble the Blue Letters


KG: You know, I'm taking guesses.

BSL: Yeah, of course, we're all taking guseess, I won’t hold you to it, don’t worry.

KG: I think that the rtaeily is, we were kind of miiontnneg before about the challenges of scale. And when you invest tens of billions of dollars in something, you need scale. And I think that's one of -- the way that AI is developed and specifically even the types of mdoels we're using, the economic model of it, which is effectively the more compute you have, the better models you can create. The better models you can cretae, the more usage you get. The more uagse you get, the better. So it has somewhat of a, honestly, like monopolistic tendency, I think, in the way that actually even like the aructihtreces and the economy of it works. And so I think it's almost inevitable that whatever AI systems are produced by these large organizations will be pushed to slace as quickly as possible. And there's some pluses in that where like, you know, sure, they're building in fdaecbek loops, people can give their input, it biases it. But also at the same time, what does it mean when a single model is fit to a billion people, right? So that's kind of what I meant about the converging effect where, what happens when we are pushed to pdoucre something that fits to a biillon people? There's a lot of diversity in there. And so, you know, we create these scaled systems that are fitting with the whole, like, trying to fit the whole plnaet. Does that work? And so I think what will, you know, we're going to go through this phase where like, yeah, you're going to have a billion pelpoe interacting the same AI. And I don't know what the effect of that will be. Even the monetization models now are kind of you pay to use these kinds of things, which are maybe OK, but certainly ads will probably enter the eaiqtoun. Also, what happens when you want attention and AI is much better at that than the algorithms you even have on YouTube and Instagram. And you can start to capture that attention. And so I certainly think it's going to be an interesting little bit here now, as we see these huge organizations spending tens of billions of dollars and the choices that they make to then monetize that, and what that means for how AI pfatlioeerrs. I know a lot of the folks in the organizations, and their interests have never been in that doaimn. But at the same time, you're beholden, you know, to stock market interests and whatever it is, then what happens? It shifts it, right? We're in a clitisapat wolrd. And that's kind of like, you know, what ultimately will change the incentives. So yeah it's interesting.

Open Cloze


KG: You know, I'm taking guesses.

BSL: Yeah, of course, we're all taking _______, I won’t hold you to it, don’t worry.

KG: I think that the _______ is, we were kind of __________ before about the challenges of scale. And when you invest tens of billions of dollars in something, you need scale. And I think that's one of -- the way that AI is developed and specifically even the types of ______ we're using, the economic model of it, which is effectively the more compute you have, the better models you can create. The better models you can ______, the more usage you get. The more _____ you get, the better. So it has somewhat of a, honestly, like monopolistic tendency, I think, in the way that actually even like the _____________ and the economy of it works. And so I think it's almost inevitable that whatever AI systems are produced by these large organizations will be pushed to _____ as quickly as possible. And there's some pluses in that where like, you know, sure, they're building in ________ loops, people can give their input, it biases it. But also at the same time, what does it mean when a single model is fit to a billion people, right? So that's kind of what I meant about the converging effect where, what happens when we are pushed to _______ something that fits to a _______ people? There's a lot of diversity in there. And so, you know, we create these scaled systems that are fitting with the whole, like, trying to fit the whole ______. Does that work? And so I think what will, you know, we're going to go through this phase where like, yeah, you're going to have a billion ______ interacting the same AI. And I don't know what the effect of that will be. Even the monetization models now are kind of you pay to use these kinds of things, which are maybe OK, but certainly ads will probably enter the ________. Also, what happens when you want attention and AI is much better at that than the algorithms you even have on YouTube and Instagram. And you can start to capture that attention. And so I certainly think it's going to be an interesting little bit here now, as we see these huge organizations spending tens of billions of dollars and the choices that they make to then monetize that, and what that means for how AI ____________. I know a lot of the folks in the organizations, and their interests have never been in that ______. But at the same time, you're beholden, you know, to stock market interests and whatever it is, then what happens? It shifts it, right? We're in a __________ _____. And that's kind of like, you know, what ultimately will change the incentives. So yeah it's interesting.

Solution


  1. reality
  2. architectures
  3. people
  4. scale
  5. create
  6. capitalist
  7. feedback
  8. planet
  9. world
  10. billion
  11. usage
  12. domain
  13. mentioning
  14. produce
  15. models
  16. guesses
  17. proliferates
  18. equation

Original Text


KG: You know, I'm taking guesses.

BSL: Yeah, of course, we're all taking guesses, I won’t hold you to it, don’t worry.

KG: I think that the reality is, we were kind of mentioning before about the challenges of scale. And when you invest tens of billions of dollars in something, you need scale. And I think that's one of -- the way that AI is developed and specifically even the types of models we're using, the economic model of it, which is effectively the more compute you have, the better models you can create. The better models you can create, the more usage you get. The more usage you get, the better. So it has somewhat of a, honestly, like monopolistic tendency, I think, in the way that actually even like the architectures and the economy of it works. And so I think it's almost inevitable that whatever AI systems are produced by these large organizations will be pushed to scale as quickly as possible. And there's some pluses in that where like, you know, sure, they're building in feedback loops, people can give their input, it biases it. But also at the same time, what does it mean when a single model is fit to a billion people, right? So that's kind of what I meant about the converging effect where, what happens when we are pushed to produce something that fits to a billion people? There's a lot of diversity in there. And so, you know, we create these scaled systems that are fitting with the whole, like, trying to fit the whole planet. Does that work? And so I think what will, you know, we're going to go through this phase where like, yeah, you're going to have a billion people interacting the same AI. And I don't know what the effect of that will be. Even the monetization models now are kind of you pay to use these kinds of things, which are maybe OK, but certainly ads will probably enter the equation. Also, what happens when you want attention and AI is much better at that than the algorithms you even have on YouTube and Instagram. And you can start to capture that attention. And so I certainly think it's going to be an interesting little bit here now, as we see these huge organizations spending tens of billions of dollars and the choices that they make to then monetize that, and what that means for how AI proliferates. I know a lot of the folks in the organizations, and their interests have never been in that domain. But at the same time, you're beholden, you know, to stock market interests and whatever it is, then what happens? It shifts it, right? We're in a capitalist world. And that's kind of like, you know, what ultimately will change the incentives. So yeah it's interesting.

Frequently Occurring Word Combinations


ngrams of length 2

collocation frequency
huge invention 5
people interact 4
immersive social 3
mental health 3
individual social 2
social experiences 2
singular world 2
shared experience 2
ai systems 2
people talk 2
social media 2
human connection 2
literally break 2
solitary confinement 2
personal achievement 2

ngrams of length 3

collocation frequency
individual social experiences 2

Important Words


  1. ads
  2. ai
  3. algorithms
  4. architectures
  5. attention
  6. beholden
  7. biases
  8. billion
  9. billions
  10. bit
  11. building
  12. capitalist
  13. capture
  14. challenges
  15. change
  16. choices
  17. compute
  18. converging
  19. create
  20. developed
  21. diversity
  22. dollars
  23. domain
  24. economic
  25. economy
  26. effect
  27. effectively
  28. enter
  29. equation
  30. feedback
  31. fit
  32. fits
  33. fitting
  34. folks
  35. give
  36. guesses
  37. hold
  38. honestly
  39. huge
  40. incentives
  41. inevitable
  42. input
  43. instagram
  44. interacting
  45. interesting
  46. interests
  47. invest
  48. kind
  49. kinds
  50. large
  51. loops
  52. lot
  53. market
  54. means
  55. meant
  56. mentioning
  57. model
  58. models
  59. monetization
  60. monetize
  61. monopolistic
  62. organizations
  63. pay
  64. people
  65. phase
  66. planet
  67. pluses
  68. produce
  69. produced
  70. proliferates
  71. pushed
  72. quickly
  73. reality
  74. scale
  75. scaled
  76. shifts
  77. single
  78. specifically
  79. spending
  80. start
  81. stock
  82. systems
  83. tendency
  84. tens
  85. time
  86. types
  87. ultimately
  88. usage
  89. work
  90. works
  91. world
  92. worry
  93. yeah
  94. youtube