vidtheque.

read-only demo

Videos zZsTVBXcbow

How Google DeepMind is researching the next Frontier of AI for Gemini — Raia Hadsell, VP of Research

index_state ready data_status ok

AI Engineer· published 2026-04-18· 0:20:37· en-US· indexed 2026-08-10 19:46

Open on YouTube

Scene timeline

  1. Shot 0, 0:00 to 0:05, 1 of 1 keyframes kept
  2. Shot 1, 0:05 to 0:09, 1 of 1 keyframes kept
  3. Shot 2, 0:09 to 0:15, 1 of 1 keyframes kept
  4. Shot 3, 0:15 to 0:26, 1 of 1 keyframes kept
  5. Shot 4, 0:26 to 0:28, 1 of 1 keyframes kept
  6. Shot 5, 0:28 to 0:34, 1 of 1 keyframes kept
  7. Shot 6, 0:34 to 0:35, 1 of 1 keyframes kept
  8. Shot 7, 0:35 to 0:38, 1 of 1 keyframes kept
  9. Shot 8, 0:38 to 0:42, 1 of 1 keyframes kept
  10. Shot 9, 0:42 to 0:44, 1 of 1 keyframes kept
  11. Shot 10, 0:44 to 0:47, 1 of 1 keyframes kept
  12. Shot 11, 0:47 to 1:04, 1 of 1 keyframes kept
  13. Shot 12, 1:04 to 1:24, 1 of 1 keyframes kept
  14. Shot 13, 1:24 to 1:26, 1 of 1 keyframes kept
  15. Shot 14, 1:26 to 2:10, 1 of 1 keyframes kept
  16. Shot 15, 2:10 to 2:40, 1 of 1 keyframes kept
  17. Shot 16, 2:40 to 3:10, 1 of 1 keyframes kept
  18. Shot 17, 3:10 to 3:35, 1 of 1 keyframes kept
  19. Shot 18, 3:35 to 4:00, 0 of 1 keyframes kept
  20. Shot 19, 4:00 to 4:26, 0 of 1 keyframes kept
  21. Shot 20, 4:26 to 4:39, 1 of 1 keyframes kept
  22. Shot 21, 4:39 to 4:47, 1 of 1 keyframes kept
  23. Shot 22, 4:47 to 4:50, 1 of 1 keyframes kept
  24. Shot 23, 4:50 to 5:03, 1 of 1 keyframes kept
  25. Shot 24, 5:03 to 5:19, 0 of 1 keyframes kept
  26. Shot 25, 5:19 to 5:46, 1 of 1 keyframes kept
  27. Shot 26, 5:46 to 6:13, 1 of 1 keyframes kept
  28. Shot 27, 6:13 to 6:40, 0 of 1 keyframes kept
  29. Shot 28, 6:40 to 7:07, 0 of 1 keyframes kept
  30. Shot 29, 7:07 to 7:34, 1 of 1 keyframes kept
  31. Shot 30, 7:34 to 8:01, 1 of 1 keyframes kept
  32. Shot 31, 8:01 to 8:30, 1 of 1 keyframes kept
  33. Shot 32, 8:30 to 8:59, 0 of 1 keyframes kept
  34. Shot 33, 8:59 to 9:25, 1 of 1 keyframes kept
  35. Shot 34, 9:25 to 9:28, 1 of 1 keyframes kept
  36. Shot 35, 9:28 to 9:50, 1 of 1 keyframes kept
  37. Shot 36, 9:50 to 10:00, 0 of 1 keyframes kept
  38. Shot 37, 10:00 to 10:27, 1 of 1 keyframes kept
  39. Shot 38, 10:27 to 10:54, 1 of 1 keyframes kept
  40. Shot 39, 10:54 to 11:41, 1 of 1 keyframes kept
  41. Shot 40, 11:41 to 12:12, 1 of 1 keyframes kept
  42. Shot 41, 12:12 to 12:42, 0 of 1 keyframes kept
  43. Shot 42, 12:42 to 13:12, 1 of 1 keyframes kept
  44. Shot 43, 13:12 to 13:23, 1 of 1 keyframes kept
  45. Shot 44, 13:23 to 13:40, 1 of 1 keyframes kept
  46. Shot 45, 13:40 to 14:06, 1 of 1 keyframes kept
  47. Shot 46, 14:06 to 14:32, 1 of 1 keyframes kept
  48. Shot 47, 14:32 to 14:43, 1 of 1 keyframes kept
  49. Shot 48, 14:43 to 15:11, 1 of 1 keyframes kept
  50. Shot 49, 15:11 to 15:18, 1 of 1 keyframes kept
  51. Shot 50, 15:18 to 15:23, 0 of 1 keyframes kept
  52. Shot 51, 15:23 to 15:24, 1 of 1 keyframes kept
  53. Shot 52, 15:24 to 15:31, 1 of 1 keyframes kept
  54. Shot 53, 15:31 to 15:32, 0 of 1 keyframes kept
  55. Shot 54, 15:32 to 15:34, 1 of 1 keyframes kept
  56. Shot 55, 15:34 to 15:35, 1 of 1 keyframes kept
  57. Shot 56, 15:35 to 15:39, 1 of 1 keyframes kept
  58. Shot 57, 15:39 to 15:41, 1 of 1 keyframes kept
  59. Shot 58, 15:41 to 15:50, 1 of 1 keyframes kept
  60. Shot 59, 15:50 to 15:51, 1 of 1 keyframes kept
  61. Shot 60, 15:51 to 15:57, 1 of 1 keyframes kept
  62. Shot 61, 15:57 to 16:07, 1 of 1 keyframes kept
  63. Shot 62, 16:07 to 16:16, 0 of 1 keyframes kept
  64. Shot 63, 16:16 to 16:25, 1 of 1 keyframes kept
  65. Shot 64, 16:25 to 16:31, 0 of 1 keyframes kept
  66. Shot 65, 16:31 to 16:36, 1 of 1 keyframes kept
  67. Shot 66, 16:36 to 16:38, 1 of 1 keyframes kept
  68. Shot 67, 16:38 to 16:43, 1 of 1 keyframes kept
  69. Shot 68, 16:43 to 16:50, 1 of 1 keyframes kept
  70. Shot 69, 16:50 to 16:52, 1 of 1 keyframes kept
  71. Shot 70, 16:52 to 16:56, 1 of 1 keyframes kept
  72. Shot 71, 16:56 to 16:58, 1 of 1 keyframes kept
  73. Shot 72, 16:58 to 17:15, 1 of 1 keyframes kept
  74. Shot 73, 17:15 to 17:21, 1 of 1 keyframes kept
  75. Shot 74, 17:21 to 17:27, 1 of 1 keyframes kept
  76. Shot 75, 17:27 to 17:29, 1 of 1 keyframes kept
  77. Shot 76, 17:29 to 17:38, 1 of 1 keyframes kept
  78. Shot 77, 17:38 to 17:39, 1 of 1 keyframes kept
  79. Shot 78, 17:39 to 17:50, 1 of 1 keyframes kept
  80. Shot 79, 17:50 to 17:51, 1 of 1 keyframes kept
  81. Shot 80, 17:51 to 17:52, 1 of 1 keyframes kept
  82. Shot 81, 17:52 to 17:59, 1 of 1 keyframes kept
  83. Shot 82, 17:59 to 18:03, 1 of 1 keyframes kept
  84. Shot 83, 18:03 to 18:04, 1 of 1 keyframes kept
  85. Shot 84, 18:04 to 18:06, 1 of 1 keyframes kept
  86. Shot 85, 18:06 to 18:09, 1 of 1 keyframes kept
  87. Shot 86, 18:09 to 18:10, 1 of 1 keyframes kept
  88. Shot 87, 18:10 to 18:13, 1 of 1 keyframes kept
  89. Shot 88, 18:13 to 18:15, 1 of 1 keyframes kept
  90. Shot 89, 18:15 to 18:22, 1 of 1 keyframes kept
  91. Shot 90, 18:22 to 18:25, 1 of 1 keyframes kept
  92. Shot 91, 18:25 to 18:29, 1 of 1 keyframes kept
  93. Shot 92, 18:29 to 18:40, 1 of 1 keyframes kept
  94. Shot 93, 18:40 to 18:41, 1 of 1 keyframes kept
  95. Shot 94, 18:41 to 18:42, 1 of 1 keyframes kept
  96. Shot 95, 18:42 to 18:44, 1 of 1 keyframes kept
  97. Shot 96, 18:44 to 18:48, 1 of 1 keyframes kept
  98. Shot 97, 18:48 to 18:56, 1 of 1 keyframes kept
  99. Shot 98, 18:56 to 19:08, 1 of 1 keyframes kept
  100. Shot 99, 19:08 to 19:11, 1 of 1 keyframes kept
  101. Shot 100, 19:11 to 19:18, 1 of 1 keyframes kept
  102. Shot 101, 19:18 to 19:20, 0 of 1 keyframes kept
  103. Shot 102, 19:20 to 19:27, 1 of 1 keyframes kept
  104. Shot 103, 19:27 to 19:30, 0 of 1 keyframes kept
  105. Shot 104, 19:30 to 19:36, 1 of 1 keyframes kept
  106. Shot 105, 19:36 to 19:44, 1 of 1 keyframes kept
  107. Shot 106, 19:44 to 19:45, 1 of 1 keyframes kept
  108. Shot 107, 19:45 to 19:48, 1 of 1 keyframes kept
  109. Shot 108, 19:48 to 20:09, 1 of 1 keyframes kept
  110. Shot 109, 20:09 to 20:22, 1 of 1 keyframes kept
  111. Shot 110, 20:22 to 20:36, 1 of 1 keyframes kept
  112. Shot 111, 20:36 to 20:36, 0 of 1 keyframes kept

112 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
194
whisperx 194
chunks
36
from 194 cues
keyframes
97
kept of 112 captured
frames with text
97
2,049 lines read
chapters
7
from the source metadata
keyframe bytes
16.2 MB
word timings on 194 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-10 01:34 2m 02s
stt done — 2026-08-10 01:36 20s
chunk done — 2026-08-10 01:36 0s
text_embed done — 2026-08-10 19:46 0s
keyframe done — 2026-08-10 01:36 2m 23s
ocr done — 2026-08-10 01:39 44s
frame_embed done — 2026-08-10 19:46 17s

Frames, and what the machine read

  • 0:03 #0 done2 line(s)

    shot 0·sharpness 666.5

    1. AlEngineer0.97
    2. EUROPE1.00

  • 0:07 #1 done2 line(s)

    shot 1·sharpness 822.1

    1. PRESENTINGSPONSOR1.00
    2. Google DeepMind1.00

  • 0:13 #2 done3 line(s)

    shot 2·sharpness 905.1

    1. PLATINUM SPONSORS0.98
    2. # Braintrust0.95
    3. WorkOS OpenAl0.94

  • 0:25 #3 done3 line(s)

    shot 3·sharpness 785.0

    1. HADSELL1.00
    2. VP OF RESEARCH0.99
    3. Google DeepMind0.98

  • 0:28 #4 done6 line(s)

    shot 4·sharpness 627.7

    1. RAIA1.00
    2. HADSELL1.00
    3. AIE1.00
    4. VP OF RESEARCH1.00
    5. Google1.00
    6. AIE1.00

  • 0:29 #5 done6 line(s)

    shot 5·sharpness 627.3

    1. RAIA1.00
    2. HADSELL1.00
    3. AIE1.00
    4. VP OF RESEARCH1.00
    5. Google1.00
    6. AIE1.00

  • 0:34 #6 done18 line(s)

    shot 6·sharpness 995.7

    1. BADSELe0.91
    2. VP OSEARCH0.96
    3. Gomagd0.61
    4. epMirad0.66
    5. meo4fj0.68
    6. SENTTRY0.88
    7. Micossoft0.80
    8. AAEE0.87
    9. AlEngineer0.99
    10. Mesdl0.51
    11. EUROPE1.00
    12. AAFEngineer0.91
    13. PRESENTED BY1.00
    14. Google DeepMind1.00
    15. ENCORD0.89
    16. Sonarr0.97
    17. :ttilscale0.91
    18. AAFE0.99

  • 0:36 #7 done25 line(s)

    shot 7·sharpness 866.0

    1. AIEngineer0.99
    2. together.ai1.00
    3. AlEngineer0.99
    4. EUROPE1.00
    5. EUROPE1.00
    6. AlEngineer0.98
    7. SAFE1.00
    8. AlEngineer0.98
    9. Braintrust1.00
    10. EUROPE1.00
    11. INTELLIGENCE1.00
    12. EUROPE1.00
    13. arize0.92
    14. AlEngineer0.97
    15. EUROPE1.00
    16. AlEngineer0.98
    17. EUROPE1.00
    18. PRESENTED BY0.96
    19. AlEngineer0.99
    20. Trigger.dev1.00
    21. Google DeepMind1.00
    22. EUROPE1.00
    23. Modal1.00
    24. AlEngineer0.99
    25. EUROPE1.00

  • 0:41 #8 done26 line(s)

    shot 8·sharpness 935.7

    1. AIEngineer0.98
    2. together.ai1.00
    3. AlEngineer0.98
    4. EUROPE1.00
    5. EUROPE1.00
    6. SAFE1.00
    7. AlEng0.97
    8. Braintrust1.00
    9. AlEngineer0.96
    10. INTELLIGENCE1.00
    11. EURI0.89
    12. EUROPE1.00
    13. AlEngineer0.98
    14. Microso0.99
    15. EUROPE1.00
    16. AIEngineer0.95
    17. EUROPE1.00
    18. PRESENTED BY1.00
    19. AI0.54
    20. Trigger.dev1.00
    21. AlEngineer0.98
    22. Google DeepMind1.00
    23. EUROPE0.96
    24. AlEngineer0.99
    25. tailscal0.97
    26. EUROPE1.00

  • 0:43 #9 done26 line(s)

    shot 9·sharpness 924.2

    1. AIEngineer0.97
    2. together.ai1.00
    3. AlEngineer0.98
    4. EUROPE1.00
    5. EUROPE1.00
    6. SAFE1.00
    7. AlEngineer0.96
    8. Braintrust1.00
    9. AIE0.89
    10. INTELLIGENCE1.00
    11. EUROPE1.00
    12. arize1.00
    13. AlEngineer0.99
    14. N0.99
    15. EUROPE1.00
    16. AIEngineer0.96
    17. EUROPE1.00
    18. PRESENTED BY1.00
    19. ngineer1.00
    20. Trigger.dev1.00
    21. AIE0.99
    22. Google DeepMind1.00
    23. EUROPE1.00
    24. Modal1.00
    25. AlEngineer0.99
    26. EUROPE1.00

  • 0:46 #10 done21 line(s)

    shot 10·sharpness 1340.9

    1. ngineer1.00
    2. togethe1.00
    3. EUROPE1.00
    4. SAF0.84
    5. AlEngineer0.98
    6. ELLIGE1.00
    7. EUROPE1.00
    8. Frontier Al and0.99
    9. the Future of Intelligence1.00
    10. ariz1.00
    11. AlEngineer0.99
    12. EUROPE1.00
    13. PRESENTED BY1.00
    14. AlEngineer1.00
    15. Google DeepMind1.00
    16. EUROPE1.00
    17. Moc1.00
    18. Raia Hadsell1.00
    19. VP, Research0.99
    20. Google DeepMind1.00
    21. Google DeepMind1.00

  • 0:58 #11 done27 line(s)

    shot 11·sharpness 2287.2

    1. gineer1.00
    2. WorkO0.99
    3. JROPE0.98
    4. AFE1.00
    5. AlEngineer0.99
    6. Frontier Al and0.99
    7. LLIGENCE1.00
    8. EUROPE1.00
    9. the Future of Intelligence1.00
    10. arize1.00
    11. AlEngineer0.99
    12. EUROPE1.00
    13. PRESENTED BY0.99
    14. IEngineer0.99
    15. Google DeepMind0.97
    16. EUROPE1.00
    17. Mod1.00
    18. Raia Hadsell1.00
    19. VP, Research0.98
    20. Google DeepMind0.99
    21. AlEngineer0.98
    22. FRONTIER AI AND THE FUTURE OF INTELLIGENCE1.00
    23. EUROPE1.00
    24. PRESENTED BY1.00
    25. Google DeepMind1.00
    26. RAIA HADSELL / VP of Research0.98
    27. Google DeepMind1.00

  • 1:18 #12 done14 line(s)

    shot 12·sharpness 2021.5

    1. Frontier Al and0.99
    2. JleDeep0.92
    3. the Future of Intelligence1.00
    4. neel0.98
    5. Raia Hadsell0.99
    6. VP,Research0.99
    7. Google DeepMind0.99
    8. AlEngineer0.98
    9. FRONTIER AI AND THE FUTURE OF INTELLIGENCE1.00
    10. EUROPE1.00
    11. PRESENTED BY1.00
    12. Google DeepMind1.00
    13. RAIA HADSELL / VP of Research0.99
    14. Google DeepMind1.00

  • 1:25 #13 done15 line(s)

    shot 13·sharpness 1843.8

    1. docs.google.com - To exit full screen, press0.98
    2. JleDeep0.87
    3. Enginee1.00
    4. EUROPE0.91
    5. 1990's1.00
    6. 2000's1.00
    7. 2010's1.00
    8. 2020's1.00
    9. AlEngineer0.98
    10. FRONTIER AI AND THE FUTURE OF INTELLIGENCE1.00
    11. EUROPE1.00
    12. PRESENTED BY1.00
    13. Google DeepMind1.00
    14. Google DeepMind1.00
    15. RAIA HADSELL / VP of Research0.98

  • 1:43 #14 done26 line(s)

    shot 14·sharpness 2061.2

    1. ineer1.00
    2. WorkOs0.93
    3. PE0.98
    4. FE1.00
    5. AlEngineer0.97
    6. LIGENCE0.99
    7. EUROPE1.00
    8. arize0.99
    9. AlEngineer0.97
    10. EUROPE1.00
    11. PRESENTED BY0.99
    12. Engineer1.00
    13. ogle DeepMind0.98
    14. EUROPE1.00
    15. Moda1.00
    16. 1990's1.00
    17. 2000's1.00
    18. 2010's1.00
    19. 2020's1.00
    20. AlEngineer0.98
    21. FRONTIER AI AND THE FUTURE OF INTELLIGENCE1.00
    22. EUROPE1.00
    23. PRESENTED BY1.00
    24. Google DeepMind1.00
    25. RAIA HADSELL / VP of Research0.99
    26. Google DeepMind1.00

  • 2:33 #15 done24 line(s)

    shot 15·sharpness 2067.9

    1. EUROPE1.00
    2. SAFE1.00
    3. AIEngine0.94
    4. INTELLIGENCE1.00
    5. EUROPE1.00
    6. ari:1.00
    7. AlEngineer0.98
    8. EUROPE1.00
    9. PRESENTED BY0.99
    10. AlEngine0.96
    11. Google DeepMind1.00
    12. EUROPE1.00
    13. Mc1.00
    14. 1990's1.00
    15. 2000's1.00
    16. 2010's1.00
    17. 2020's1.00
    18. AlEngineer0.98
    19. FRONTIER AI AND THE FUTURE OF INTELLIGENCE0.99
    20. EUROPE1.00
    21. PRESENTED BY1.00
    22. Google DeepMind1.00
    23. RAIA HADSELL / VP of Research0.99
    24. Google DeepMind1.00

  • 2:49 #16 done20 line(s)

    shot 16·sharpness 1872.3

    1. gineer1.00
    2. WorkC0.99
    3. ROPE1.00
    4. O0.74
    5. AIEngineer0.95
    6. EUROPE0.98
    7. ginee0.98
    8. JROPE0.94
    9. 1990's1.00
    10. 2000's1.00
    11. 2010's1.00
    12. 2020's1.00
    13. ilsca1.00
    14. AlEngineer0.99
    15. FRONTIER AI AND THE FUTURE OF INTELLIGENCE1.00
    16. EUROPE1.00
    17. PRESENTED BY1.00
    18. Google DeepMind1.00
    19. Google DeepMind1.00
    20. RAIA HADSELL / VP of Research0.98

  • 3:32 #17 done26 line(s)

    shot 17·sharpness 3165.6

    1. EUROPE1.00
    2. DeepMind Frontier Al1.00
    3. runp0.97
    4. AIEngin0.93
    5. EUROPE1.00
    6. Our Mission:0.99
    7. Create the future of intelligence0.99
    8. Trigge1.00
    9. • Find the root nodes0.99
    10. AlEngineer0.97
    11. • Partner with the world0.99
    12. EUROPE1.00
    13. PRESENTED BY1.00
    14. AIEngin0.95
    15. • Solve problems that are worth0.97
    16. Google DeepMind0.99
    17. EUROPE1.00
    18. solving1.00
    19. CLOU1.00
    20. AlEngineer0.98
    21. FRONTIER AI AND THE FUTURE OF INTELLIGENCE1.00
    22. EUROPE1.00
    23. PRESENTED BY1.00
    24. Google DeepMind1.00
    25. RAIA HADSELL / VP of Research0.97
    26. Google DeepMind1.00

  • 3:38 #18 skipped

    shot 18·duplicate of #17

  • 4:20 #19 skipped

    shot 19·duplicate of #17

  • 4:34 #20 done17 line(s)

    shot 20·sharpness 2836.0

    1. DeepMind Frontier Al1.00
    2. DS0.84
    3. Our Mission:0.97
    4. Create the future of intelligence0.99
    5. • Find the root nodes0.97
    6. ● Partner with the world0.96
    7. ● Solve problems that are worth0.97
    8. solving1.00
    9. Goc0.84
    10. AlEngineer0.98
    11. FRONTIER AI AND THE FUTURE OF INTELLIGENCE1.00
    12. EUROPE1.00
    13. PRESENTED BY1.00
    14. Google DeepMind1.00
    15. # Braintrust0.94
    16. WorkOS1.00
    17. OpenAl0.93

  • 4:41 #21 done10 line(s)

    shot 21·sharpness 2106.4

    1. DeepMind Frontier Al0.97
    2. Our Mission:0.99
    3. Create the future of intelligence0.99
    4. • Find the root nodes0.96
    5. ● Partner with the world0.97
    6. ● Solve problems that are worth0.99
    7. solving1.00
    8. Goc0.77
    9. Engineering the future of Al0.99
    10. AlEngineer0.99

  • 4:49 #22 done36 line(s)

    shot 22·sharpness 2145.2

    1. IEnginer0.94
    2. Brain1.00
    3. EUROPE1.00
    4. Agentic0.97
    5. Worlds1.00
    6. Al for Humans0.98
    7. Sustainability0.98
    8. pei0.92
    9. 1Engin0.92
    10. and general embodied1.00
    11. Advance world models0.99
    12. agents.1.00
    13. Human-Alcollabiration.0.96
    14. socialintelligence,health0.96
    15. and education.0.98
    16. Modelling weather.0.98
    17. energy, and planet earth.1.00
    18. EUROPE0.94
    19. Engin1.00
    20. EUROPE1.00
    21. Creative1.00
    22. Advanced1.00
    23. Future of1.00
    24. Mo1.00
    25. Technologies1.00
    26. Push the frontiers of0.99
    27. Al-empowered creativity.1.00
    28. systems and advance0.98
    29. multimodal research.1.00
    30. Models1.00
    31. Design new learning1.00
    32. Solve AGl in the physical1.00
    33. Robotics1.00
    34. world.0.96
    35. Engineering the future of Al0.99
    36. AlEngineer0.99

  • 4:53 #23 done39 line(s)

    shot 23·sharpness 2319.6

    1. EUROPE1.00
    2. Tess1.00
    3. AIEngin0.96
    4. EUROPE1.00
    5. Sno0.92
    6. Agentic1.00
    7. Worlds1.00
    8. Al for Humans0.97
    9. Sustainability1.00
    10. AlEngineer0.96
    11. EUROPE1.00
    12. and general embodied0.99
    13. Advance world models1.00
    14. agents.1.00
    15. Human-Alcollabiration.0.96
    16. social intelligence, health0.99
    17. and education.0.97
    18. Modelling weather.0.98
    19. energy, and planet earth.1.00
    20. PRESENTED BY1.00
    21. AIEngin0.94
    22. Google DeepMind0.99
    23. EUROPE1.00
    24. so0.93
    25. Creative1.00
    26. Advanced1.00
    27. Future of1.00
    28. Technologies1.00
    29. Models1.00
    30. Robotics1.00
    31. Push the frontiers of1.00
    32. Al-empowered creativity.0.99
    33. systems and advance1.00
    34. multimodal research.1.00
    35. Design new learning0.99
    36. Solve AGl in the physical0.99
    37. world.1.00
    38. Engineering the future of Al0.99
    39. AlEngineer0.99

Transcript

194 cues· 2,796 words· 15,048 chars

  1. 0:14 Our next speaker is VP of Research at Google DeepMind.
  2. 0:19 Please join me in welcoming to the stage Raya Hasdel.
  3. 0:34 Hello, everyone.
  4. 0:35 Wonderful.
  5. 0:37 What a lovely full room and good smiles.
  6. 0:39 I heard the dig on Google there at the end.
  7. 0:42 I did catch that.
  8. 0:45 So my name is Raya Hadsell.
  9. 0:47 I've been a part of DeepMind for the last almost 13 years.
  10. 0:54 And I'm very happy to have AI Engineer come here to London.
  11. 0:58 I'm also very proud this year to be
  12. 1:02 a UK AI ambassador, so I help the government, academia, industry sort of bridge those gaps.
  13. 1:11 And yes, I'm American by birth, but I've been here for long enough that I can count myself among the proud Brits as well.
  14. 1:19 So I'm going to talk a little bit about frontier AI and the future of intelligence.
  15. 1:26 to start a little bit of a longer introduction to who I am.
  16. 1:30 It's good to be as old as I am.
  17. 1:33 You get to look at this by the decades.
  18. 1:35 So in the 90s, I did my undergraduate degree in philosophy of religion.
  19. 1:40 I was definitely not a computer scientist yet, but I really enjoyed it, before you ask.
  20. 1:51 Yes, I learned a lot, and I'm glad that I did it, and no, it hasn't been very useful since.
  21. 1:57 In the 2000s, I did a bit of a pivot, moved into computer science after some good advice from those close to me, and spent my PhD years in New York City working on convolutional networks, neural networks for robots with Yann LeCun, which is a lot of fun.
  22. 2:20 I then, in the 2010s, made the decision to join a small group of curious, scrappy individuals working at DeepMind.
  23. 2:32 It's a group of about 30, 40 people at the time.
  24. 2:36 And we spent the rest of that decade working on things like Atari video games, Go, StarCraft, and some robotics.
  25. 2:47 A lot of fun.
  26. 2:49 Now I am a VP within DeepMind.
  27. 2:54 I help run a group of about 1,200 scientists and engineers across 10 labs.
  28. 3:04 And we're working on a lot of different things.
  29. 3:07 I'll tell you about three of those.
  30. 3:11 So first, frontier AI is an area where we really are trying to make sure that we are staying in the front.
  31. 3:21 So we're thinking about what are the next architectures that we're going to use for Gemini?
  32. 3:27 What are the next problems that only AI can really address?
  33. 3:32 And how are we going to build the future of intelligence?
  34. 3:37 And that's thinking not just about artificial intelligence, but it's create the future of human intelligence as well, and even robotics intelligence as well.
  35. 3:47 We are all on this journey together, and I think that it's important to think about how humans change as well as the technology.
  36. 3:57 Our approach, we look for root nodes.
  37. 3:59 We're not gonna waste time on the leaves.
  38. 4:01 We're gonna really find for a big problem space that hasn't been solved, how deep can we go?
  39. 4:08 Find the deepest problems and solve those in order to then enable a lot of downstream impact.
  40. 4:18 We partner really with the world.
  41. 4:20 I really think about it very broadly and think about who are the partners that can help us find those root nodes and solve those problems and also bring it to the leaf nodes and solving problems that are worth solving.
  42. 4:34 The motto or the mission of DeepMind is to build AI responsibly for the benefit of humanity.
  43. 4:41 So I really take that seriously.
  44. 4:43 We want to solve problems that are worth solving.
  45. 4:48 All right, so we work in a lot of different areas within Frontier AI.
  46. 4:54 In DeepMind, these are sort of some of the different categories.
  47. 4:58 I'm not going to tell you about all of them, so you can just maybe keep those a mystery.
  48. 5:04 But I'll just pick out a couple.
  49. 5:05 So first, in advanced models, I actually wanted to bring up an embeddings model.
  50. 5:13 So the theme of this talk overall is things that are not directly language models.

Chapters

  1. 0:00 Introduction
  2. 5:05 Advanced Embedding Models: Raia discusses the importance of embedding models for fast retrieval and recognition, similar to how the human brain uses 'Jennifer Aniston cells' to identify concepts across modalities. She highlights Gemini Embeddings 2, a fully omnimodal model that processes text, video, and audio into unified semantic vectors.
  3. 9:53 AI for Weather Forecasting: The team has developed revolutionary models for atmospheric prediction, moving away from traditional physics simulations. Notable breakthroughs include
  4. 11:00 GraphCast: A spherical graph neural network that provides accurate 15-day weather forecasts.
  5. 12:47 GenCast: A probabilistic model that offers higher efficiency and accuracy (97% of the time compared to gold-standard benchmarks).
  6. 13:51 FGN: A functional generative network that directly predicts cyclone behavior, which is currently being utilized by the US National Hurricane Center.
  7. 14:35 World Models: Hadsell introduces Genie, a project focused on creating interactive, real-time environments. Starting from Genie 1 (2D platformers) and progressing to Genie 3, these models allow users to create and interact with high-quality, 3D photorealistic worlds. These environments demonstrate capabilities like memory, consistency, and the ability to be dynamically prompted by the user to change the surroundings in real-time.

Open at this second