WEBVTT
1
00:00:03.200 --> 00:00:17.320
Calaruga, Shark Media. November seventeenth, twenty twenty three, twelve twenty
2
00:00:17.359 --> 00:00:21.399
eight pm Pacific time. Sam Altman is in a Las
3
00:00:21.480 --> 00:00:25.359
Vegas hotel room, preparing for the APEX summit, where he'll
4
00:00:25.359 --> 00:00:30.359
represent American AI innovation to world leaders. His phone buzzes
5
00:00:30.399 --> 00:00:34.039
with a Google meet link from Ilia Suitskaver, open AI's
6
00:00:34.119 --> 00:00:37.880
chief scientist and his co founder. Sam assumes it's about
7
00:00:37.920 --> 00:00:43.000
the new research Ilia's been excited about. Twelve thirty pm.
8
00:00:43.439 --> 00:00:47.560
Sam joins the call. The entire board is there, unusual
9
00:00:47.600 --> 00:00:51.920
for a Friday afternoon. Ilia's face looks stricken like he's
10
00:00:51.960 --> 00:00:57.200
about to deliver a death diagnosis. In a way he is, sam,
11
00:00:57.520 --> 00:01:02.320
Ilia says, his voice shaking. You're being terminated as CEO
12
00:01:02.920 --> 00:01:08.599
effective immediately. The words hang in digital space. Sam Altman,
13
00:01:09.120 --> 00:01:12.280
the face of the AI revolution, the man who turned
14
00:01:12.400 --> 00:01:15.920
chat GPT into a verb, has just been fired from
15
00:01:15.959 --> 00:01:21.280
the company he created. No warning, no performance review, no
16
00:01:21.480 --> 00:01:26.120
chance to defend himself. You haven't been consistently candid with
17
00:01:26.200 --> 00:01:29.879
the board. Helen Toner, adds, the independent board member who's
18
00:01:29.879 --> 00:01:34.439
been increasingly vocal about AI safety. We've lost confidence in
19
00:01:34.480 --> 00:01:37.799
your ability to lead open AI. The call ends at
20
00:01:37.799 --> 00:01:41.439
twelve thirty nine p m. Nine minutes that will nearly
21
00:01:41.480 --> 00:01:45.439
destroy a ninety billion dollar company, send shock waves through
22
00:01:45.480 --> 00:01:49.719
Silicon Valley, and turn Thanksgiving Week two thousand twenty three
23
00:01:50.079 --> 00:01:55.120
into the tech industry's most dramatic corporate coup. Within five days,
24
00:01:55.519 --> 00:01:59.400
ninety five percent of open AI's employees will threaten to quit,
25
00:02:00.439 --> 00:02:03.719
Microsoft will offer to hire them all, the board will
26
00:02:03.719 --> 00:02:07.480
beg for mercy, and Sam Altman will orchestrate one of
27
00:02:07.480 --> 00:02:13.840
the greatest comebacks in corporate history. Welcome back to feuds, power, pride,
28
00:02:13.919 --> 00:02:18.599
and payback. I'm Marcus Reid. Some feuds are about money,
29
00:02:19.039 --> 00:02:22.680
others are about power, but the rarest feuds are about
30
00:02:22.680 --> 00:02:26.599
the future of humanity itself, or at least that's what
31
00:02:26.680 --> 00:02:30.319
both sides believed when open AI's board went to war
32
00:02:30.439 --> 00:02:34.719
with its CEO over who gets to control artificial intelligence.
33
00:02:35.520 --> 00:02:38.919
This week marks two years since the Thanksgiving coup that
34
00:02:39.000 --> 00:02:43.479
almost killed chat GPT, destroyed a ninety billion dollar company,
35
00:02:43.879 --> 00:02:48.639
and proved that in Silicon Valley, employee loyalty trump's board authority.
36
00:02:48.759 --> 00:02:54.120
Every single time, it's a story of idealism versus capitalism,
37
00:02:54.639 --> 00:02:58.719
safety versus acceleration, and what happens when the people building
38
00:02:58.800 --> 00:03:02.080
god like AI can't even agree on who should lead
39
00:03:02.120 --> 00:03:07.960
their company. This is episode thirty seven, The Five Days
40
00:03:08.000 --> 00:03:17.159
That Almost Killed Ai, open AI's Thanksgiving Massacre, Act one,
41
00:03:17.879 --> 00:03:22.479
the Missionaries and the Mercenary. To understand how OpenAI went
42
00:03:22.520 --> 00:03:26.840
from humanity's hope to corporate chaos in nine minutes, you
43
00:03:27.000 --> 00:03:30.800
need to understand the fundamental contradiction built into its DNA
44
00:03:31.199 --> 00:03:35.560
from day one. Open Ai was founded in December twenty
45
00:03:35.680 --> 00:03:41.280
fifteen as a nonprofit with a messianic mission ensure artificial
46
00:03:41.319 --> 00:03:46.080
general intelligence benefits all of humanity. The founding team read
47
00:03:46.120 --> 00:03:50.759
like a who's who of tech royalty, Elon Musk, Sam Altman,
48
00:03:51.199 --> 00:03:56.080
Ilia Sutskiver, Greg Brockman, and others. United by the belief
49
00:03:56.120 --> 00:03:59.439
that AI was too powerful to be left to Google
50
00:03:59.599 --> 00:04:03.360
or face Book. We wanted to make sure AGI didn't
51
00:04:03.439 --> 00:04:07.319
end up controlled by a single corporation, Greg Brockman, open
52
00:04:07.360 --> 00:04:13.360
AI's president, explained, this was about saving humanity, not making money.
53
00:04:13.759 --> 00:04:17.720
The irony of billionaires creating a nonprofit to save humanity
54
00:04:18.279 --> 00:04:23.360
wasn't lost on anyone, but the mission attracted true believers.
55
00:04:24.120 --> 00:04:28.160
Chief among them was Elia Suitskiver, a soft spoken genius
56
00:04:28.439 --> 00:04:32.279
who'd studied under Jeffrey Hinton, the godfather of deep learning.
57
00:04:33.319 --> 00:04:36.560
Ilia wasn't interested in money, he already had plenty from
58
00:04:36.639 --> 00:04:41.360
Google's acquisition of DNN Research. He wanted to solve intelligence itself.
59
00:04:42.120 --> 00:04:46.399
Sam Altman was a different creature entirely. The Stanford dropout
60
00:04:46.399 --> 00:04:51.800
had run Y Combinator, Silicon Valley's most prestigious startup accelerator,
61
00:04:52.160 --> 00:04:57.920
with ruthless efficiency. He was brilliant, ambitious, and understood something
62
00:04:57.959 --> 00:05:02.639
the researchers didn't. But intentions don't pay for GPU clusters
63
00:05:02.879 --> 00:05:07.560
that cost millions per month. Sam was always the business
64
00:05:07.560 --> 00:05:11.519
guy among the priests. A former open ai employee told me,
65
00:05:12.160 --> 00:05:15.680
we were building a new religion. He was building a company.
66
00:05:16.600 --> 00:05:20.560
The first major crack came in twenty nineteen, when Altman
67
00:05:20.720 --> 00:05:26.319
orchestrated OpenAI's transformation from pure nonprofit to a capped profit
68
00:05:26.399 --> 00:05:31.399
structure of Frankenstein's Monster of corporate forms that would allow
69
00:05:31.439 --> 00:05:37.759
investment while theoretically maintaining the mission. Elon Musk quit in disgust,
70
00:05:38.319 --> 00:05:42.160
calling it a betrayal of the founding principles. But Altman
71
00:05:42.399 --> 00:05:46.319
was just getting started. He convinced Microsoft to invest one
72
00:05:46.439 --> 00:05:51.399
billion dollars, then ten billion dollars. More, he pushed researchers
73
00:05:51.399 --> 00:05:56.959
to release products, not just papers. When GPT three showed promise,
74
00:05:57.360 --> 00:06:01.199
he wanted it in user's hands immediately. When chat GPT
75
00:06:01.399 --> 00:06:05.439
exploded in November twenty twenty two, becoming the fastest growing
76
00:06:05.480 --> 00:06:11.519
application in history, Altman was everywhere Congress conferences, cable news
77
00:06:11.920 --> 00:06:16.120
as the face of AI. Sam loved being the spokesman
78
00:06:16.199 --> 00:06:19.800
for the future, another insider said, but some of us
79
00:06:19.839 --> 00:06:23.519
worried he was speaking for a future he didn't fully understand.
80
00:06:24.600 --> 00:06:29.120
The board watched with growing alarm. Helen Toner, a director
81
00:06:29.120 --> 00:06:33.519
at Georgetown's Center for Security and Emerging Technology, had joined
82
00:06:33.519 --> 00:06:38.000
the board as an AI safety expert. She published research
83
00:06:38.120 --> 00:06:42.600
suggesting Open AI was moving too fast, releasing powerful systems
84
00:06:42.600 --> 00:06:47.360
without adequate safeguards. Tasha mc caulay, another board member, shared
85
00:06:47.399 --> 00:06:53.680
these concerns, but their biggest ally was unexpected. Ilya Suitskever himself,
86
00:06:54.560 --> 00:06:58.079
the chief scientist who'd helped build chat GPT, was having
87
00:06:58.120 --> 00:07:02.319
second thoughts about what they'd unleashed. Ilia went from being
88
00:07:02.399 --> 00:07:07.399
Sam's biggest supporter to his biggest worry. A researcher close
89
00:07:07.480 --> 00:07:11.480
to both men revealed he'd see these demos of GPT
90
00:07:11.759 --> 00:07:16.879
four and later models and literally lose sleep. He kept saying,
91
00:07:17.319 --> 00:07:22.720
We're not ready, Humanity's not ready. By fall twenty twenty three,
92
00:07:23.120 --> 00:07:26.920
the board was holding secret meetings without Sam. The topic
93
00:07:27.720 --> 00:07:32.079
was there CEO putting growth before safety? Was he being
94
00:07:32.480 --> 00:07:37.519
consistently candid about open AI's capabilities and his own ambitions?
95
00:07:38.360 --> 00:07:41.639
Was Sam Altman the right person to shepherd humanity into
96
00:07:41.680 --> 00:07:46.639
the age of artificial intelligence? On November sixteenth, they reached
97
00:07:46.680 --> 00:07:59.720
their verdict Act two, the Thanksgiving Massacre, the execution of
98
00:07:59.720 --> 00:08:02.920
the who was as brutal as it was bungled. After
99
00:08:02.959 --> 00:08:07.160
firing Sam via video call, the board immediately called Greg Brockman,
100
00:08:07.560 --> 00:08:13.519
open AI's president and Sam's closest ally. Greg, we've made
101
00:08:13.560 --> 00:08:17.199
a change in leadership, Adam DiAngelo, the Corus CEO and
102
00:08:17.319 --> 00:08:22.279
board member, said carefully, Sam is no longer CEO, but
103
00:08:22.360 --> 00:08:26.000
we want you to stay on. Greg's response was immediate,
104
00:08:26.480 --> 00:08:30.399
are you fucking kidding me? Within minutes, Greg had texted
105
00:08:30.439 --> 00:08:34.879
Sam quit his position and started rallying the troops. The
106
00:08:34.960 --> 00:08:39.879
board somehow hadn't anticipated that firing the beloved CEO might
107
00:08:40.000 --> 00:08:44.480
upset his equally beloved co founder. At three point thirty pm,
108
00:08:44.759 --> 00:08:48.039
OpenAI published a blog post that would go down in
109
00:08:48.200 --> 00:08:53.799
history as one of tech's worst crisis communications. Mister Altman's
110
00:08:53.799 --> 00:08:57.960
departure follows a deliberative review process by the board, which
111
00:08:58.000 --> 00:09:01.519
concluded that he was not consistently candid in his communications
112
00:09:01.679 --> 00:09:05.960
with the board, hindering its ability to exercise its responsibilities.
113
00:09:06.679 --> 00:09:09.120
The board no longer has confidence in his ability to
114
00:09:09.120 --> 00:09:14.279
continue leading open ai. The vagueness was intentional and catastrophic.
115
00:09:14.960 --> 00:09:19.759
By not specifying Sam's alleged transgressions, they let imagination run wild.
116
00:09:20.600 --> 00:09:25.360
Had he lied about AI capabilities, hidden a breakthrough, committed fraud.
117
00:09:26.279 --> 00:09:30.279
The ambiguity created a vacuum that sucked in every conspiracy
118
00:09:30.320 --> 00:09:36.399
theory imaginable. Microsoft open AI's biggest investor and partner, found
119
00:09:36.440 --> 00:09:40.960
out about the firing from the public announcement. Satya Nadella,
120
00:09:41.399 --> 00:09:45.679
Microsoft CEO, was in Seattle preparing for a celebration of
121
00:09:45.720 --> 00:09:51.759
their partnership. Instead, his phone exploded with calls from panicked executives.
122
00:09:52.480 --> 00:09:57.519
Their thirteen billion dollar investment, Their entire AI strategy was
123
00:09:57.600 --> 00:10:03.159
suddenly in jeopardy. Satya went from confused to furious to
124
00:10:03.200 --> 00:10:08.840
strategic in about thirty minutes. A Microsoft insider said by dinner,
125
00:10:09.320 --> 00:10:12.879
he was already working on plan B. Hire everyone from
126
00:10:12.919 --> 00:10:18.519
open ai and rebuild it inside Microsoft. At open AI's
127
00:10:18.559 --> 00:10:24.200
San Francisco headquarters, chaos reigned. Employees learned their CEO had
128
00:10:24.240 --> 00:10:29.240
been fired from Twitter, Slack channels, erupted with disbelief, anger,
129
00:10:29.639 --> 00:10:34.120
and rapidly organizing resistance. The board had somehow forgotten that
130
00:10:34.200 --> 00:10:39.600
in tech employees have unprecedented power, especially employees who could
131
00:10:39.639 --> 00:10:42.879
walk across the street and get hired for double their salary.
132
00:10:43.879 --> 00:10:47.399
By Friday evening, the narrative had shifted from board fire's
133
00:10:47.480 --> 00:10:52.879
CEO to board stages coup Sam. Ever, the strategist wasn't
134
00:10:52.919 --> 00:10:58.320
fighting back yet. He tweeted simply, I love open Ai
135
00:10:58.440 --> 00:11:00.879
and everything I've done has been in service of keeping
136
00:11:00.960 --> 00:11:04.559
it true to its mission. Then came his master stroke.
137
00:11:05.159 --> 00:11:08.240
On Saturday, he posted a photo of himself at open
138
00:11:08.279 --> 00:11:13.639
AI's offices wearing a guest badge captioned first and last
139
00:11:13.759 --> 00:11:18.039
day as open Ai visitor. The image went viral, the
140
00:11:18.120 --> 00:11:21.879
fired CEO returning like a visiting dignitary to the company.
141
00:11:21.919 --> 00:11:26.440
He built, radiating confidence while the board hid from public view.
142
00:11:27.399 --> 00:11:31.919
The employee revolt began in earnest. By Sunday, morning, over
143
00:11:32.000 --> 00:11:35.879
five hundred of Open ais seven hundred seventy employees had
144
00:11:35.919 --> 00:11:39.919
signed an open letter with a simple message, bring back
145
00:11:40.000 --> 00:11:52.679
Sam and Greg where we quit? Act three the counter revolution.
146
00:11:53.440 --> 00:11:57.480
Elia Suitscaver cracked first. The chief scientist who delivered the
147
00:11:57.519 --> 00:12:00.279
killing blow, was having second thoughts. By Saturday, Monie warning
148
00:12:01.120 --> 00:12:05.519
friends reported he hadn't slept, was barely eating, kept checking
149
00:12:05.600 --> 00:12:09.399
the employee letter as name after name of his researchers appeared.
150
00:12:11.000 --> 00:12:15.679
Ilia realized he'd been played. A close colleague revealed the
151
00:12:15.720 --> 00:12:19.360
board used his concerns about safety to justify their power grab.
152
00:12:19.919 --> 00:12:23.639
But this wasn't making AI safer. It was destroying the
153
00:12:23.799 --> 00:12:29.360
very institution meant to develop it responsibly. On Sunday afternoon,
154
00:12:29.799 --> 00:12:33.840
Ilia posted the tweet that signaled the board's doom. I
155
00:12:34.039 --> 00:12:38.159
deeply regret my participation in the board's actions. I never
156
00:12:38.200 --> 00:12:42.240
intended to harm Open AI. I love everything we've built together,
157
00:12:42.720 --> 00:12:45.840
and I will do everything I can to reunite the company.
158
00:12:47.320 --> 00:12:52.480
The chief scientist had switched sides. The coup was collapsing. Meanwhile,
159
00:12:52.799 --> 00:12:56.559
the board was desperately trying to find a new CEO.
160
00:12:56.679 --> 00:13:01.120
They approached Nat Friedman, former GitHub CEO. OH he said no,
161
00:13:02.039 --> 00:13:08.440
they begged Miramadi, open AI's CTO, who they'd named interim CEO,
162
00:13:09.240 --> 00:13:13.320
she made it clear she wanted Sam back. In desperation,
163
00:13:13.759 --> 00:13:17.519
they turned to Emmett Sheer, Twitch's co founder, who agreed
164
00:13:17.559 --> 00:13:20.480
to take the role late Sunday night, but it was
165
00:13:20.519 --> 00:13:24.799
too late. Satya Nadella had made his move. At one
166
00:13:24.919 --> 00:13:28.919
am Monday morning, he announced that Microsoft was hiring Sam
167
00:13:28.919 --> 00:13:32.559
Altman and Greg Brockman to lead a new advanced AI
168
00:13:32.679 --> 00:13:38.000
research team. The kicker, any open ai employee who wanted
169
00:13:38.000 --> 00:13:41.960
to join them was welcome. It was checkmate. A board
170
00:13:41.960 --> 00:13:46.440
adviser later admitted Satya basically said, you can keep the
171
00:13:46.519 --> 00:13:50.240
name open ai, but we'll take everything else, the people,
172
00:13:50.679 --> 00:13:56.519
the technology, the momentum. By Monday afternoon, the employee letter
173
00:13:56.559 --> 00:14:01.559
had over seven hundred signatures, including Ilia Suitskiver. The board
174
00:14:01.600 --> 00:14:05.080
faced a stark reality. They could oversee an empty building
175
00:14:05.120 --> 00:14:09.039
with the open ai name, or they could surrender. Helen
176
00:14:09.120 --> 00:14:13.159
Toner and Tasha McCauley wanted to fight on, convinced they
177
00:14:13.159 --> 00:14:18.519
were saving humanity from reckless AI development, but Adam DiAngelo,
178
00:14:18.759 --> 00:14:23.159
the pragmatist, saw the writing on the wall. Open AI
179
00:14:23.320 --> 00:14:27.960
without its people was worthless. The board's fiduciary duty, even
180
00:14:28.000 --> 00:14:32.320
at a nonprofit, was to preserve the organization. That meant
181
00:14:32.399 --> 00:14:37.240
bringing Sam back. The negotiations Monday night into Tuesday were brutal.
182
00:14:37.799 --> 00:14:41.919
Sam had all the leverage and used it his demands.
183
00:14:41.960 --> 00:14:46.360
Helen Toner and Tasha McCauley off the board immediately. Greg
184
00:14:46.440 --> 00:14:52.039
Brockman reinstated an independent investigation into the board's actions, and,
185
00:14:52.120 --> 00:14:55.840
most importantly, a new board structure that would prevent this
186
00:14:55.879 --> 00:15:09.639
from ever happening again. Act for the King returns at
187
00:15:09.679 --> 00:15:12.799
one and Wednesday morning, five days after the coup began,
188
00:15:13.320 --> 00:15:18.360
OpenAI announced Sam Altman's return as CEO. The staff celebrated
189
00:15:18.399 --> 00:15:21.679
like their city had been liberated. Photos from the office
190
00:15:21.679 --> 00:15:26.200
showed employees crying, hugging, and holding up signs reading open
191
00:15:26.240 --> 00:15:30.879
ai is nothing without its people. Sam's return tweet was
192
00:15:31.000 --> 00:15:35.960
gracious but pointed. I love open ai and everything I've
193
00:15:36.000 --> 00:15:38.879
done over the past few days has been in service
194
00:15:38.919 --> 00:15:43.080
of keeping this team and its mission together. I'm looking
195
00:15:43.120 --> 00:15:46.039
forward to returning to open ai and building on our
196
00:15:46.080 --> 00:15:52.799
strong partnership with Microsoft. Translation, I won, you lost, and
197
00:15:52.919 --> 00:15:56.919
Microsoft is now explicitly our daddy. The new board was
198
00:15:56.919 --> 00:16:03.200
announced Brett Taylor, former Salesforce co CEO as chair, Larry Summers,
199
00:16:03.559 --> 00:16:08.600
former Treasury secretary, and Adam DiAngelo, the only survivor from
200
00:16:08.600 --> 00:16:13.240
the old board. Helen Toner and Tasha macauley were gone,
201
00:16:13.279 --> 00:16:16.679
their concerns about AI safety dismissed as the rantings of
202
00:16:16.759 --> 00:16:21.320
Luddites who tried to stop progress. In the aftermath, the
203
00:16:21.360 --> 00:16:25.000
real reasons for the coup slowly emerged. Yes, there were
204
00:16:25.080 --> 00:16:30.200
legitimate concerns about AI safety and Sam's communication style, but
205
00:16:30.240 --> 00:16:33.960
there were also whispers about Sam's side projects, a chip
206
00:16:34.000 --> 00:16:37.919
start up, an energy company that the board felt represented
207
00:16:38.039 --> 00:16:42.480
conflicts of interest. There were concerns about Sam accumulating too
208
00:16:42.559 --> 00:16:47.360
much power, becoming the singular face of AI, but none
209
00:16:47.399 --> 00:16:50.919
of that mattered now. The employees had voted with their feet.
210
00:16:51.600 --> 00:16:55.639
Microsoft had shown who really controlled Open AI, and Sam
211
00:16:55.679 --> 00:17:00.159
Altman had proven that in Silicon Valley, founder CEOs are
212
00:17:00.320 --> 00:17:04.200
kings who can only be deposed if their subjects allow it.
213
00:17:11.599 --> 00:17:16.559
Epilogue The Cost of Revolution Two years later, as we
214
00:17:16.640 --> 00:17:20.799
record this episode in November twenty twenty five, the aftermath
215
00:17:20.839 --> 00:17:24.359
of the Thanksgiving massacre has reshaped not just Open AI
216
00:17:24.759 --> 00:17:30.759
but the entire AI industry. Sam Altman emerged more powerful
217
00:17:30.799 --> 00:17:34.599
than ever. The company that was almost destroyed is now
218
00:17:34.680 --> 00:17:39.759
Silicon Valley's most valuable startup. Sam tours the world like
219
00:17:39.799 --> 00:17:43.279
a head of state, meeting with presidents and prime ministers,
220
00:17:43.799 --> 00:17:47.880
shaping AI policy globally. The board members who tried to
221
00:17:47.920 --> 00:17:52.400
stop him have been vindicated in some ways. AI capabilities
222
00:17:52.599 --> 00:17:56.160
are advancing at a pace that terrifies even some researchers,
223
00:17:56.880 --> 00:18:01.160
but they've been thoroughly repudiated in their methods. Helen Toner
224
00:18:01.240 --> 00:18:05.960
left to focus on AI policy research, Tasha mcaulay retreated
225
00:18:05.960 --> 00:18:09.960
from public view. Their concerns about safety were real, but
226
00:18:10.079 --> 00:18:13.160
their coup attempt made the martyrs for a cause that
227
00:18:13.279 --> 00:18:18.279
now seems quaint. Ilia Suitskaver, the brilliant researcher caught in
228
00:18:18.319 --> 00:18:21.640
the middle, left open Ai in May twenty twenty four
229
00:18:22.079 --> 00:18:26.480
to start his own company focused on safe superintelligence. His
230
00:18:26.599 --> 00:18:30.799
departure was friendly, but everyone knows the real reason he
231
00:18:30.799 --> 00:18:34.160
couldn't stay at the company after betraying his friend, even
232
00:18:34.200 --> 00:18:38.640
if that betrayal lasted less than forty eight hours. Microsoft's
233
00:18:38.680 --> 00:18:42.960
position has been formalized and strengthened. They're no longer just
234
00:18:43.000 --> 00:18:49.160
an investor, but open AI's explicit guardian Satia Nadella's swift
235
00:18:49.200 --> 00:18:52.599
action during the crisis proved that when push comes to shove,
236
00:18:53.240 --> 00:18:57.519
the three trillion dollar tech giant calls the shots. The
237
00:18:57.519 --> 00:19:01.839
broader lesson for Silicon Valley was clear boards serve at
238
00:19:01.880 --> 00:19:05.759
the pleasure of founders and employees, not the other way around.
239
00:19:06.680 --> 00:19:10.039
The old model of independent board oversight died in those
240
00:19:10.079 --> 00:19:14.200
five days in November. Every startup board member now knows
241
00:19:14.240 --> 00:19:18.920
that crossing a popular CEO means war with the entire company,
242
00:19:19.720 --> 00:19:23.839
But the deeper questions remain unanswered. Was the board right
243
00:19:23.920 --> 00:19:27.640
to worry about AI safety? Even if their methods were wrong?