Understanding AI from Predicting to Acting — Transcript

Full transcript

1. 0:00In 1957, a scientist at the National
2. 0:02Bureau of Standards scanned a photograph
3. 0:04of his baby son into a computer 176 by
4. 0:07176 pixels, barely visible by today's
5. 0:09standards. It was the first digital
6. 0:11image ever created. That single pixel
7. 0:13eventually led to everything from
8. 0:15Instagram to AI-generated art. The same
9. 0:18digitization we talked about earlier is
10. 0:20now being supercharged by algorithms
11. 0:22that can learn, predict, and create. So,
12. 0:24what is artificial intelligence or AI
13. 0:26actually? Let's break it down. You may
14. 0:28have heard the terms weak AI and strong
15. 0:31AI, a [music] way of separating machines
16. 0:33built to do one narrow task from a
17. 0:36hypothetical machine with human-like
18. 0:38general intelligence. [music] That
19. 0:40distinction used to be clean. It no
20. 0:41longer is. Today's systems are far more
21. 0:45general purpose than the weak AI label
22. 0:47suggests, but they still fall well short
23. 0:50of true general intelligence or [music]
24. 0:52artificial general intelligence, AGI.
25. 0:55Instead of a simple two-category split,
26. 0:58I want to give you a more useful
27. 1:00framework, one based on what these
28. 1:02systems actually do. Predictive AI,
29. 1:05generative AI, and agentic AI. The first
30. 1:08and oldest category is predictive AI.
31. 1:11These are systems trained to recognize
32. 1:13patterns in data and forecast or
33. 1:15classify what comes next.
34. 1:16>> [music]
35. 1:17>> Think about the last time a platform
36. 1:18showed you a video you never searched
37. 1:20for and you couldn't stop watching. That
38. 1:22wasn't a coincidence. YouTube's
39. 1:24recommendation algorithm drives over 70%
40. 1:26of all watch time on the platform.
41. 1:29>> [music]
42. 1:29>> Not what you searched for, but what the
43. 1:31system predicted you'd want next.
44. 1:33Netflix does the same. So does the fraud
45. 1:36detection behind your banking app. At
46. 1:39its core, AI labels examples, finds the
47. 1:43patterns, predicts the outcome.
48. 1:45>> [music]
49. 1:45>> This is the AI you've been using for
50. 1:47years, even if you called it the
51. 1:50algorithm. The second category is
52. 1:52generative AI, and this is where most of
53. 1:54the recent excitement comes from.
54. 1:56Instead of just predicting an outcome,
55. 1:58these systems create something new,
56. 2:00text, images, code, music. Foundation
57. 2:03models like GPT-4, Claude, and Gemini
58. 2:05are trained on massive amounts of data
59. 2:07and can write, analyze, summarize, and
60. 2:09generate, often from a single
61. 2:11conversation. This is genuinely
62. 2:13different from predictive AI. A
63. 2:16recommendation engine tells you what
64. 2:18you'll probably like. A generative model
65. 2:20can write the article,
66. 2:22>> [music]
67. 2:22>> draft the code, or produce the image
68. 2:24itself. While it's not AGI because it
69. 2:27doesn't reason the way a person does
70. 2:29across every domain,
71. 2:31it's far beyond a single-task tool. The
72. 2:34third category is agentic AI, and it's
73. 2:37the one moving fastest right now. This
74. 2:40isn't hypothetical. Agentic systems
75. 2:42already exist and are being deployed
76. 2:45today.
77. 2:46An agentic system doesn't just answer a
78. 2:48question or generate one output. It can
79. 2:51take a goal, break it into steps, use
80. 2:53tools, and carry out a complex task with
81. 2:56limited human input along the way.
82. 2:59Imagine asking an AI to plan a business
83. 3:01trip. Instead of just suggesting
84. 3:03flights, an agentic system can compare
85. 3:05options, check your calendar for
86. 3:06conflicts, book the ticket, and confirm
87. 3:08the itinerary with you at the end, not
88. 3:11at every single step. The same logic
89. 3:13applies to a customer service request,
90. 3:15not just drafting a reply, but checking
91. 3:17an order, issuing a refund, and closing
92. 3:19the ticket. Adoption of these agentic
93. 3:21systems that act, not just respond, is
94. 3:23accelerating quickly. Whenever I
95. 3:25describe these capabilities, someone
96. 3:27asks the same question, does this
97. 3:28replace people? Sometimes in narrow
98. 3:30ways, yes, but the more important story
99. 3:33is augmentation, how AI changes what a
100. 3:35task, a role, or an organization can do,
101. 3:37rather than simply eliminating the
102. 3:39person doing it. Consider a dentist
103. 3:41using an AI system to help read X-rays.
104. 3:44The goal isn't to remove the dentist,
105. 3:46it's to catch what a tired or rushed
106. 3:48human eye might miss, so the dentist's
107. 3:50judgment gets sharper, not obsolete.
108. 3:52>> [music]
109. 3:52>> Rather than replacing the workforce, AI
110. 3:55will necessarily change how the people
111. 3:56and teams work. We've seen this pattern
112. 3:58when any new technology or process is
113. 4:00introduced. That pattern will keep
114. 4:02repeating. With AI, these changes are
115. 4:04happening quickly.
116. 4:05Three categories evolving fast.
117. 4:08Predictive AI forecasts and classifies.
118. 4:10[music]
119. 4:12Generative AI creates something new.
120. 4:14Agentic AI takes action across multiple
121. 4:17steps toward a goal. And all three are
122. 4:20real and in active use today. Together,
123. 4:24predictive, [music] generative, and
124. 4:25agentic AI are already reshaping tasks,
125. 4:29roles, and organizations. Not by
126. 4:31replacing people wholesale, but by
127. 4:33changing what's possible for the people
128. 4:36still in the loop. [music] As you think
129. 4:37about your future work, ask which of
130. 4:40these three could actually strengthen
131. 4:42what I'm building.
132. 4:43In the next video, we'll look at how
133. 4:45large organizations are investing in AI
134. 4:48and the importance of leadership and
135. 4:49successful absorption of AI and workflow
136. 4:51redesign.