如何高效合并音视频文件（时间短消耗资源少）（二）

英语字幕

1
00:00:06,480 --> 00:00:08,400
Good morning. We have a banger for you2
00:00:08,400 --> 00:00:09,840
today. We're going to launch chatbt3
00:00:09,840 --> 00:00:11,519
agent. But before jumping into that, I'd4
00:00:11,519 --> 00:00:12,559
like to ask the team to introduce5
00:00:12,559 --> 00:00:14,080
themselves. Starting with Yosh.6
00:00:14,080 --> 00:00:17,840
Hi, I'm Yash. I work on agent team and7
00:00:17,840 --> 00:00:20,080
before that I used to work on operator.8
00:00:20,080 --> 00:00:22,560
Hi, I'm Jing. I work on agents research9
00:00:22,560 --> 00:00:24,400
previously on deep research.10
00:00:24,400 --> 00:00:26,000
Hi, I'm Casey. I'm a researcher on11
00:00:26,000 --> 00:00:27,920
agents formerly operator.12
00:00:27,920 --> 00:00:30,560
Hi, I'm Issa. I'm a researcher on agent13
00:00:30,560 --> 00:00:32,640
formerly on deep research.14
00:00:32,640 --> 00:00:34,880
So we we started launching agents15
00:00:34,880 --> 00:00:36,800
earlier this year. Uh we launched deep16
00:00:36,800 --> 00:00:38,879
research, we launched operator and17
00:00:38,879 --> 00:00:40,160
people were very excited about this.18
00:00:40,160 --> 00:00:42,480
People could see that now uh AI was19
00:00:42,480 --> 00:00:44,640
going off to do complex tasks for them.20
00:00:44,640 --> 00:00:46,079
But it became clear to us that what21
00:00:46,079 --> 00:00:48,000
people really wanted was for us to bring22
00:00:48,000 --> 00:00:49,760
those capabilities and more together.23
00:00:49,760 --> 00:00:51,920
People wanted a unified agent that could24
00:00:51,920 --> 00:00:55,039
go off, use its own computer and do real25
00:00:55,039 --> 00:00:57,360
complex tasks for them, that could uh26
00:00:57,360 --> 00:00:59,359
seamlessly transition from thinking27
00:00:59,359 --> 00:01:01,520
about something to taking actions to28
00:01:01,520 --> 00:01:03,359
using lots of tools using the terminal,29
00:01:03,359 --> 00:01:05,360
clicking around the web, even producing30
00:01:05,360 --> 00:01:06,880
things like spreadsheets and slides and31
00:01:06,880 --> 00:01:08,960
and much more. And wanted people want to32
00:01:08,960 --> 00:01:10,159
be able to do this over a long time33
00:01:10,159 --> 00:01:12,159
horizon and a sort of for universal34
00:01:12,159 --> 00:01:13,840
tasks. So the team has been working35
00:01:13,840 --> 00:01:16,400
super hard to bring that together. And36
00:01:16,400 --> 00:01:18,080
today we have chat with the agent. Um,37
00:01:18,080 --> 00:01:19,680
it's probably easier to show it to you38
00:01:19,680 --> 00:01:21,439
than to keep talking about it. It is one39
00:01:21,439 --> 00:01:23,360
of the feel the aon moments for me to40
00:01:23,360 --> 00:01:25,280
watch it work. So, let's take a look.41
00:01:25,280 --> 00:01:27,840
Awesome. Thanks, Sam. Hello, everyone.42
00:01:27,840 --> 00:01:29,920
Very excited to share chat GBD agent43
00:01:29,920 --> 00:01:31,600
with everybody. And as Sam said, let's44
00:01:31,600 --> 00:01:33,759
just dive right into the demo. Okay, so45
00:01:33,759 --> 00:01:36,159
we are on Chad GBD as we all know and46
00:01:36,159 --> 00:01:39,119
love. And to turn on the agent mode, you47
00:01:39,119 --> 00:01:40,880
just click the tools menu and select48
00:01:40,880 --> 00:01:43,280
agent. You can also just type agent in49
00:01:43,280 --> 00:01:45,040
the composer bar and it'll take you to50
00:01:45,040 --> 00:01:47,520
agent mode. Um, Edward and I have a51
00:01:47,520 --> 00:01:49,360
wedding to go to later this year. Uh,52
00:01:49,360 --> 00:01:51,119
it's for one of our mutual friends.53
00:01:51,119 --> 00:01:52,560
Should we should we have the Asian54
00:01:52,560 --> 00:01:53,280
planet?55
00:01:53,280 --> 00:01:55,680
Yeah, let's do it. I need an outfit. And56
00:01:55,680 --> 00:01:56,799
don't forget the gift.57
00:01:56,799 --> 00:01:58,719
Okay, great. We won't forget the gift.58
00:01:58,719 --> 00:02:00,240
Um, it's a little bit of a longer59
00:02:00,240 --> 00:02:01,680
prompt, so I have it copied in my60
00:02:01,680 --> 00:02:02,799
buffer, so I'm just going to go ahead61
00:02:02,799 --> 00:02:05,759
and paste it. Um, okay. So, let's see.62
00:02:05,759 --> 00:02:07,360
Let's see what it says. Our friends are63
00:02:07,360 --> 00:02:08,640
getting married later this year, as I64
00:02:08,640 --> 00:02:10,720
said, Minia and Sarah. And we want the65
00:02:10,720 --> 00:02:12,879
agent to help us find an outfit that66
00:02:12,879 --> 00:02:15,520
matches the dress code. uh propose a few67
00:02:15,520 --> 00:02:17,840
options. Nice mid luxury taking into68
00:02:17,840 --> 00:02:21,040
account venue and weather. We also want69
00:02:21,040 --> 00:02:23,280
to find us some hotels and as Edward70
00:02:23,280 --> 00:02:25,760
said, don't forget the gift. Um so let's71
00:02:25,760 --> 00:02:27,840
see and72
00:02:27,840 --> 00:02:30,319
send the prompt away. As Sam said, agent73
00:02:30,319 --> 00:02:32,640
uses a computer. Uh so in the beginning74
00:02:32,640 --> 00:02:34,959
it sets up its environment. It it you75
00:02:34,959 --> 00:02:38,000
know it'll take a minute or two or not76
00:02:38,000 --> 00:02:39,680
really 5 seconds to set up its77
00:02:39,680 --> 00:02:41,440
environment. And in this case, as you78
00:02:41,440 --> 00:02:43,840
see, it understands the prompt. It's79
00:02:43,840 --> 00:02:46,319
asking for me for a clarification. I'm80
00:02:46,319 --> 00:02:48,000
just going to let it just continue and81
00:02:48,000 --> 00:02:51,120
work. Anyway, um I think it got confused82
00:02:51,120 --> 00:02:54,239
by saying, "Oh, where's the um what83
00:02:54,239 --> 00:02:55,680
exactly is the time of the date of the84
00:02:55,680 --> 00:02:57,200
wedding?" I think it'll figure out using85
00:02:57,200 --> 00:02:59,840
the website. Okay, cool. So, now it's86
00:02:59,840 --> 00:03:01,760
kicked off. It's starting the process,87
00:03:01,760 --> 00:03:03,920
the prompt, and it's open up a browser.88
00:03:03,920 --> 00:03:04,959
And to walk you through what's89
00:03:04,959 --> 00:03:06,800
happening, here's90
00:03:06,800 --> 00:03:09,040
Yeah. So, as mentioned, we gave the91
00:03:09,040 --> 00:03:10,879
agent access to its own virtual92
00:03:10,879 --> 00:03:13,280
computer, and the computer has many93
00:03:13,280 --> 00:03:14,720
different tools installed, and it can94
00:03:14,720 --> 00:03:16,239
choose which to use as it's working95
00:03:16,239 --> 00:03:18,640
through the task. So, in chat GPT, you96
00:03:18,640 --> 00:03:21,360
can see a visualization of the agent's97
00:03:21,360 --> 00:03:23,680
computer screen, and you can see98
00:03:23,680 --> 00:03:25,519
overlaid its chain of thought in text,99
00:03:25,519 --> 00:03:27,200
and that's what it's thinking as it's100
00:03:27,200 --> 00:03:28,480
working through the task and deciding101
00:03:28,480 --> 00:03:30,799
what to do next. We gave the agent102
00:03:30,799 --> 00:03:32,400
access to two different ways to browse103
00:03:32,400 --> 00:03:34,560
the internet. First, we gave it a text104
00:03:34,560 --> 00:03:36,159
browser, and this is similar to the deep105
00:03:36,159 --> 00:03:38,000
research tool. And this is what lets it106
00:03:38,000 --> 00:03:40,159
really efficiently and quickly read many107
00:03:40,159 --> 00:03:43,440
web pages um um and search for them. And108
00:03:43,440 --> 00:03:45,040
we also gave it access to a visual109
00:03:45,040 --> 00:03:46,319
browser. And this is similar to the110
00:03:46,319 --> 00:03:48,239
operator tool. And this is what lets it111
00:03:48,239 --> 00:03:50,159
actually interact with the UI of a web112
00:03:50,159 --> 00:03:52,720
page. So it can um drag things. It can113
00:03:52,720 --> 00:03:54,879
use the cursor to click around. It can114
00:03:54,879 --> 00:03:57,280
open UI components. It can fill out115
00:03:57,280 --> 00:03:59,920
forms and enter text and text areas.116
00:03:59,920 --> 00:04:02,560
It's very flexible. So those two tools117
00:04:02,560 --> 00:04:04,720
are very complimentary. And then we also118
00:04:04,720 --> 00:04:06,720
gave it access to its own terminal so119
00:04:06,720 --> 00:04:08,720
that it can run code and it can also120
00:04:08,720 --> 00:04:10,640
generate and analyze files like slide121
00:04:10,640 --> 00:04:12,879
decks and spreadsheets. And then through122
00:04:12,879 --> 00:04:14,560
the terminal it's also able to call123
00:04:14,560 --> 00:04:17,840
APIs. So both public APIs and APIs to124
00:04:17,840 --> 00:04:19,840
access your private data sources like125
00:04:19,840 --> 00:04:22,479
Google Drive, Google Calendar, GitHub,126
00:04:22,479 --> 00:04:25,360
SharePoint and many others um and only127
00:04:25,360 --> 00:04:26,960
if you explicitly connect them similar128
00:04:26,960 --> 00:04:28,960
to deep research connectors. And then it129
00:04:28,960 --> 00:04:31,680
also has access to the image gen API so130
00:04:31,680 --> 00:04:34,240
it can create nice visuals for um slide131
00:04:34,240 --> 00:04:36,080
decks and other things as it's working132
00:04:36,080 --> 00:04:38,240
through its tasks.133
00:04:38,240 --> 00:04:40,800
How is deciding which tools to use here?134
00:04:40,800 --> 00:04:42,560
Yes, we train the model to move between135
00:04:42,560 --> 00:04:44,160
these capabilities with reinforcement136
00:04:44,160 --> 00:04:46,080
learning. This is the first model we137
00:04:46,080 --> 00:04:48,880
trained that has access to this unified138
00:04:48,880 --> 00:04:52,000
tool box. A text browser, a GUI browser139
00:04:52,000 --> 00:04:53,840
and a terminal all in one virtual140
00:04:53,840 --> 00:04:57,120
machine. To guide its learning, we141
00:04:57,120 --> 00:04:59,360
created hard tasks that require using142
00:04:59,360 --> 00:05:01,919
all these tools. This allows the model143
00:05:01,919 --> 00:05:04,000
not only to learn how to use these144
00:05:04,000 --> 00:05:06,160
tools, but also when to use which tool145
00:05:06,160 --> 00:05:08,400
depending on the task at hand. At the146
00:05:08,400 --> 00:05:10,400
beginning of the training, the model147
00:05:10,400 --> 00:05:12,880
might attempt to use all these tools to148
00:05:12,880 --> 00:05:15,600
solve a relatively simple problem. Over149
00:05:15,600 --> 00:05:17,840
time, as we reward the model for solving150
00:05:17,840 --> 00:05:20,560
problems correctly and efficiently, the151
00:05:20,560 --> 00:05:24,080
model will have smarter tool choice.152
00:05:24,080 --> 00:05:27,360
For example, if you ask a model to uh153
00:05:27,360 --> 00:05:29,039
find a restaurant with specific154
00:05:29,039 --> 00:05:31,919
requirements and make a reservation, the155
00:05:31,919 --> 00:05:34,479
model may typically just start a deep156
00:05:34,479 --> 00:05:36,160
research in the text browser to find157
00:05:36,160 --> 00:05:39,039
some candidates, then switch to the GUI158
00:05:39,039 --> 00:05:42,160
browser to view photos of food, uh check159
00:05:42,160 --> 00:05:45,600
availability, and complete the booking.160
00:05:45,600 --> 00:05:48,000
Similarly, for creative task like161
00:05:48,000 --> 00:05:50,160
creating an artifact, the model will162
00:05:50,160 --> 00:05:51,680
first search online for public163
00:05:51,680 --> 00:05:54,479
resources, then switch to the terminal164
00:05:54,479 --> 00:05:57,039
to do some code editing to compile the165
00:05:57,039 --> 00:05:59,919
artifact and finally verify the final166
00:05:59,919 --> 00:06:02,960
outputs in the GUI browser. With this,167
00:06:02,960 --> 00:06:05,600
we truly feel like we brought together168
00:06:05,600 --> 00:06:08,240
the best of deep research and operator169
00:06:08,240 --> 00:06:11,759
and added some extra sparkle.170
00:06:11,759 --> 00:06:14,000
That's right. Yeah. So to put this171
00:06:14,000 --> 00:06:15,520
project in context, I want to give a bit172
00:06:15,520 --> 00:06:18,000
of history. So a few months ago, we173
00:06:18,000 --> 00:06:20,960
shipped operator in January and this was174
00:06:20,960 --> 00:06:23,120
our agent that lets you do online tasks175
00:06:23,120 --> 00:06:25,759
like book reservations and um send176
00:06:25,759 --> 00:06:27,840
emails and then two weeks later we177
00:06:27,840 --> 00:06:29,919
shipped deep research and deep research178
00:06:29,919 --> 00:06:31,919
is a tool that lets you do in-depth179
00:06:31,919 --> 00:06:35,759
internet research and output highquality180
00:06:35,759 --> 00:06:39,280
um um research reports. And after launch181
00:06:39,280 --> 00:06:41,039
we realized that actually these two182
00:06:41,039 --> 00:06:42,319
approaches are actually deeply183
00:06:42,319 --> 00:06:44,160
complimentary.184
00:06:44,160 --> 00:06:46,400
Um for example operator has some trouble185
00:06:46,400 --> 00:06:48,720
reading super long articles. Um it has186
00:06:48,720 --> 00:06:50,400
to scroll. It takes a long time. But187
00:06:50,400 --> 00:06:51,759
that's something that deep research is188
00:06:51,759 --> 00:06:56,240
good at. Conversely operator uh uh deep189
00:06:56,240 --> 00:06:58,240
research isn't as good at interacting190
00:06:58,240 --> 00:07:00,319
with web pages interactive elements191
00:07:00,319 --> 00:07:03,199
visual uh highly visual web pages but192
00:07:03,199 --> 00:07:04,800
that's something that operator excels193
00:07:04,800 --> 00:07:08,639
at. So uh yeah we felt these approaches194
00:07:08,639 --> 00:07:11,120
were complimentary and then we we were195
00:07:11,120 --> 00:07:13,120
also looking at some customer feedback.196
00:07:13,120 --> 00:07:14,880
So for example one of our most highly197
00:07:14,880 --> 00:07:17,120
requested features for deep research was198
00:07:17,120 --> 00:07:18,960
the ability to log into websites and199
00:07:18,960 --> 00:07:20,960
access authenticated sources. That's200
00:07:20,960 --> 00:07:22,880
something that operator can do.201
00:07:22,880 --> 00:07:24,000
I've been waiting for that for a long202
00:07:24,000 --> 00:07:24,560
time.203
00:07:24,560 --> 00:07:26,160
Yeah.204
00:07:26,160 --> 00:07:28,479
Um another thing is that we were looking205
00:07:28,479 --> 00:07:29,840
at the prompts that people were trying206
00:07:29,840 --> 00:07:31,520
for operator and we saw that they were207
00:07:31,520 --> 00:07:32,880
actually more deep research type208
00:07:32,880 --> 00:07:35,199
prompts. for example, plan a trip and209
00:07:35,199 --> 00:07:38,240
then book it. And so, yeah, we we really210
00:07:38,240 --> 00:07:39,360
feel like we're bringing the best of211
00:07:39,360 --> 00:07:41,440
both worlds here. And on a personal212
00:07:41,440 --> 00:07:42,800
note, we've all been friends for a213
00:07:42,800 --> 00:07:44,160
while, and it's really exciting to be214
00:07:44,160 --> 00:07:46,479
working together. So, speaking of215
00:07:46,479 --> 00:07:48,960
matches made in heaven, how is the216
00:07:48,960 --> 00:07:50,319
wedding planning going?217
00:07:50,319 --> 00:07:51,759
It's amazing to watch. This is an218
00:07:51,759 --> 00:07:53,599
example of a task I hate doing. This can219
00:07:53,599 --> 00:07:55,520
like ruin like, you know, multiple hours220
00:07:55,520 --> 00:07:56,960
for me as I get sucked into these rabbit221
00:07:56,960 --> 00:07:58,160
holes. So, just watching this as you222
00:07:58,160 --> 00:07:59,520
guys have been talking click through223
00:07:59,520 --> 00:08:01,199
this and just like do the whole thing is224
00:08:01,199 --> 00:08:03,360
really quite remarkable. Yeah, totally.225
00:08:03,360 --> 00:08:06,560
Um, looks like it started off by226
00:08:06,560 --> 00:08:08,560
figuring out the weather. One of the227
00:08:08,560 --> 00:08:11,280
cool features, um, is that, you know, as228
00:08:11,280 --> 00:08:12,560
some of these tasks may take a little229
00:08:12,560 --> 00:08:14,160
bit longer, you can just go back and see230
00:08:14,160 --> 00:08:15,759
what it was doing. So, that's what we're231
00:08:15,759 --> 00:08:17,199
exactly going to do. Looks like it went232
00:08:17,199 --> 00:08:18,720
through the website to use the text233
00:08:18,720 --> 00:08:21,039
browser. Interestingly, for that, now234
00:08:21,039 --> 00:08:22,400
it's looking through the suits for235
00:08:22,400 --> 00:08:23,919
Edward. I think it'll find something236
00:08:23,919 --> 00:08:25,360
good. Here you can see it switched over237
00:08:25,360 --> 00:08:27,199
to actually a visual browser to make238
00:08:27,199 --> 00:08:28,960
sure suit will look really good on239
00:08:28,960 --> 00:08:31,280
Edward.240
00:08:31,280 --> 00:08:34,560
And now looks like yeah, it's got241
00:08:34,560 --> 00:08:36,880
chugging along, figuring out what to do.242
00:08:36,880 --> 00:08:39,599
Um, and still on suits and now probably243
00:08:39,599 --> 00:08:41,919
getting to the gifts section. Um, okay,244
00:08:41,919 --> 00:08:43,279
cool. So, this is going to take a while.245
00:08:43,279 --> 00:08:44,959
As Sam said, these tasks sometimes can246
00:08:44,959 --> 00:08:46,160
take a long time. So, it's going to247
00:08:46,160 --> 00:08:47,680
continue doing hopefully much faster248
00:08:47,680 --> 00:08:49,760
than we will do. Um, should we do249
00:08:49,760 --> 00:08:51,600
something else while it's doing it? I250
00:08:51,600 --> 00:08:53,519
think the team really wanted the um251
00:08:53,519 --> 00:08:55,279
stickers, some stickers for the for the252
00:08:55,279 --> 00:08:56,480
launch. Should we do that?253
00:08:56,480 --> 00:08:57,279
Yeah, cool.254
00:08:57,279 --> 00:08:59,040
All right. So, we have a team mascot,255
00:08:59,040 --> 00:09:00,320
which is one of our colleagues, Bunny256
00:09:00,320 --> 00:09:03,279
Doodle. really really cute tell you. Um257
00:09:03,279 --> 00:09:06,080
and we're going to try and bring um get258
00:09:06,080 --> 00:09:08,480
some laptop stickers for everybody. Uh259
00:09:08,480 --> 00:09:10,480
one of the favorite features for agent260
00:09:10,480 --> 00:09:13,120
is given that trajectories can take 15261
00:09:13,120 --> 00:09:15,040
minutes, 20 minutes, 30 minutes262
00:09:15,040 --> 00:09:17,120
depending on the complexity of the task.263
00:09:17,120 --> 00:09:19,120
Um a lot of times the you might need to264
00:09:19,120 --> 00:09:20,560
help the agent. Agent might need to ask265
00:09:20,560 --> 00:09:22,480
you clarifications, confirmations and266
00:09:22,480 --> 00:09:25,040
things like that. Um so I love to use it267
00:09:25,040 --> 00:09:26,640
on the go. So I'm going to use my mobile268
00:09:26,640 --> 00:09:28,160
phone to actually send the query this269
00:09:28,160 --> 00:09:30,240
time and then see how it goes.270
00:09:30,240 --> 00:09:32,880
Okay, so let's see. Okay, so we are on271
00:09:32,880 --> 00:09:35,519
Chad Gibbdi. Uh I have already selected272
00:09:35,519 --> 00:09:38,560
the agent mode. I've also inputed our uh273
00:09:38,560 --> 00:09:40,560
cute mascot and I'm going to quickly274
00:09:40,560 --> 00:09:43,040
paste a query. So query says make some275
00:09:43,040 --> 00:09:45,279
swag for the team one by one laptop276
00:09:45,279 --> 00:09:47,920
stickers and order 500 of them. I'll277
00:09:47,920 --> 00:09:52,959
also say I like sticker mule278
00:09:52,959 --> 00:09:55,279
which we have used in the past and send279
00:09:55,279 --> 00:09:57,200
it off.280
00:09:57,200 --> 00:10:00,080
Okay. So, just like it was doing on the281
00:10:00,080 --> 00:10:02,080
web, it's going to take some time, think282
00:10:02,080 --> 00:10:04,080
about like what's it doing, and it'll283
00:10:04,080 --> 00:10:07,120
kick off kick off the query. And as it's284
00:10:07,120 --> 00:10:08,880
going, it'll take some time to kick it285
00:10:08,880 --> 00:10:11,200
off. Is it Oh, there we go. So, it'll286
00:10:11,200 --> 00:10:12,480
start working on it. Looks like it's287
00:10:12,480 --> 00:10:14,720
starting to create the anime art. It'll288
00:10:14,720 --> 00:10:16,640
probably use image that Isa referred289
00:10:16,640 --> 00:10:18,399
earlier on to create hopefully an anime290
00:10:18,399 --> 00:10:20,240
art. We'll see how it comes out. While291
00:10:20,240 --> 00:10:21,760
that's going, anything else we want to292
00:10:21,760 --> 00:10:22,399
do?293
00:10:22,399 --> 00:10:24,720
Oh, yeah. I also need a pair of shoes294
00:10:24,720 --> 00:10:26,320
because my shoes got damaged.295
00:10:26,320 --> 00:10:27,360
How did they get damaged?296
00:10:27,360 --> 00:10:28,560
Uh, by the rain297
00:10:28,560 --> 00:10:30,000
in SF.298
00:10:30,000 --> 00:10:30,800
Yes.299
00:10:30,800 --> 00:10:32,160
Cool. All right. Uh, well, let's get300
00:10:32,160 --> 00:10:34,240
Edward a pair of shoes as well. So, oh,301
00:10:34,240 --> 00:10:40,320
can you also find us um pair of men's302
00:10:40,320 --> 00:10:43,519
dress black shoes in size303
00:10:43,519 --> 00:10:44,240
9.5?304
00:10:44,240 --> 00:10:46,000
9.5.305
00:10:46,000 --> 00:10:47,920
So, one of the key capabilities of the306
00:10:47,920 --> 00:10:49,920
model is being able to interrupt. I307
00:10:49,920 --> 00:10:51,920
think you know as trajectories take long308
00:10:51,920 --> 00:10:53,760
time or whatever time it's really309
00:10:53,760 --> 00:10:56,720
important for us to for it to feel very310
00:10:56,720 --> 00:10:59,120
multi-turn so the users can interject311
00:10:59,120 --> 00:11:01,120
user can direct it user can give it more312
00:11:01,120 --> 00:11:02,640
guidance less guidance whatever we want313
00:11:02,640 --> 00:11:04,320
to do and that's what we're doing here314
00:11:04,320 --> 00:11:07,040
we essentially the the model was315
00:11:07,040 --> 00:11:08,720
chugging along figuring out all the316
00:11:08,720 --> 00:11:10,240
things that we had asked before and in317
00:11:10,240 --> 00:11:12,320
this case we essentially said hey can318
00:11:12,320 --> 00:11:16,000
you also uh get us a pair of men's black319
00:11:16,000 --> 00:11:18,160
shoes and now it's thinking and soon320
00:11:18,160 --> 00:11:19,839
enough hopefully it'll take that into321
00:11:19,839 --> 00:11:22,000
account and keep going uh into its322
00:11:22,000 --> 00:11:23,600
trajectory. There we go. So, it said323
00:11:23,600 --> 00:11:25,120
acknowledge the interruption. It said,324
00:11:25,120 --> 00:11:26,880
"Okay, cool. I'll also research men's325
00:11:26,880 --> 00:11:29,600
black shoes in size 9.5." Um, and then326
00:11:29,600 --> 00:11:31,680
it'll probably get on its way. Um, but327
00:11:31,680 --> 00:11:33,120
maybe Issa can tell us a little bit more328
00:11:33,120 --> 00:11:34,240
about how that works.329
00:11:34,240 --> 00:11:36,320
Yeah, sure. So, as you can see, the330
00:11:36,320 --> 00:11:38,079
agent is very collaborative, and this331
00:11:38,079 --> 00:11:39,920
was really important to us when we were332
00:11:39,920 --> 00:11:41,200
training the model and building the333
00:11:41,200 --> 00:11:42,880
product. If you were asking another334
00:11:42,880 --> 00:11:44,399
person to do a task for you that would335
00:11:44,399 --> 00:11:45,519
take them a really long time to336
00:11:45,519 --> 00:11:46,959
complete, you'd probably give them some337
00:11:46,959 --> 00:11:48,800
instructions to start and then they338
00:11:48,800 --> 00:11:50,640
might ask you some clarifying questions339
00:11:50,640 --> 00:11:52,320
and then they'd start the task and maybe340
00:11:52,320 --> 00:11:53,600
realize, oh, they need more341
00:11:53,600 --> 00:11:55,440
clarification from you or they need your342
00:11:55,440 --> 00:11:56,880
permission to sign into something or do343
00:11:56,880 --> 00:11:58,560
something on your behalf and then you344
00:11:58,560 --> 00:12:00,240
might realize, oh, I forgot to mention345
00:12:00,240 --> 00:12:02,640
this thing or um what's your status? How346
00:12:02,640 --> 00:12:04,240
are you doing? Can I help redirect you347
00:12:04,240 --> 00:12:05,760
if you're getting along the wrong path348
00:12:05,760 --> 00:12:07,760
or something? And so similarly for these349
00:12:07,760 --> 00:12:09,680
really longrunning agentic tasks, it's350
00:12:09,680 --> 00:12:11,519
very important that both the user and351
00:12:11,519 --> 00:12:13,600
the agent are able to initiate352
00:12:13,600 --> 00:12:15,519
communication with each other so that um353
00:12:15,519 --> 00:12:17,200
the agent is able to most effectively354
00:12:17,200 --> 00:12:19,360
help you with your tasks. And so this is355
00:12:19,360 --> 00:12:20,560
something that we actually trained into356
00:12:20,560 --> 00:12:22,320
the model. We trained it to be able to357
00:12:22,320 --> 00:12:24,160
ask clarifying questions, not every358
00:12:24,160 --> 00:12:26,240
single time like deep research. Um we359
00:12:26,240 --> 00:12:28,800
also asked it we also trained it to be360
00:12:28,800 --> 00:12:30,560
interruptible as Yash just showed. And361
00:12:30,560 --> 00:12:32,000
also sometimes it will ask you for362
00:12:32,000 --> 00:12:33,519
clarification and confirmation363
00:12:33,519 --> 00:12:35,680
mid-trajectory.364
00:12:35,680 --> 00:12:38,079
Yeah. And part of working with agent is365
00:12:38,079 --> 00:12:40,480
that well sometimes it'll make mistakes.366
00:12:40,480 --> 00:12:42,079
And that's why we felt it was important367
00:12:42,079 --> 00:12:44,079
to train the model to ask you for368
00:12:44,079 --> 00:12:45,920
confirmation at the last step of369
00:12:45,920 --> 00:12:49,279
important steps. Um so for example maybe370
00:12:49,279 --> 00:12:51,519
before it's going to send the email um371
00:12:51,519 --> 00:12:53,440
it'll ask you to take a look at the372
00:12:53,440 --> 00:12:54,720
draft and whether it makes sense and373
00:12:54,720 --> 00:12:56,079
whether there are any embarrassing374
00:12:56,079 --> 00:12:59,200
typos. Um, and if there are, then you375
00:12:59,200 --> 00:13:01,360
can either ask it to fix it or you can376
00:13:01,360 --> 00:13:03,440
directly take over the browser and jump377
00:13:03,440 --> 00:13:06,079
right into the um, agents environment378
00:13:06,079 --> 00:13:09,040
and correct it yourself. And that way it379
00:13:09,040 --> 00:13:10,720
feels collaborative and you can um,380
00:13:10,720 --> 00:13:13,680
really work with the agent.381
00:13:13,680 --> 00:13:15,120
Should we look at maybe one more demo?382
00:13:15,120 --> 00:13:17,279
We've got this uh, sort of fun tradition383
00:13:17,279 --> 00:13:19,600
in live streams of using uh, using our384
00:13:19,600 --> 00:13:21,120
newest models to sort of evaluate385
00:13:21,120 --> 00:13:23,040
themselves or do something kind of meta.386
00:13:23,040 --> 00:13:24,240
Anything like that we could do?387
00:13:24,240 --> 00:13:27,440
Yeah, let's do it.388
00:13:27,440 --> 00:13:28,320
So um389
00:13:28,320 --> 00:13:29,440
I think people would love to know how390
00:13:29,440 --> 00:13:30,320
good the model is.391
00:13:30,320 --> 00:13:33,920
Yes. So this is a prompt we previously392
00:13:33,920 --> 00:13:36,880
gave the a agent yesterday. So basically393
00:13:36,880 --> 00:13:38,959
it asks the model to pull its own394
00:13:38,959 --> 00:13:40,959
evalution number from our Google job395
00:13:40,959 --> 00:13:43,440
connector and make some slides. So we396
00:13:43,440 --> 00:13:44,959
want to keep it simple like no397
00:13:44,959 --> 00:13:47,360
introduction no conclusion just present398
00:13:47,360 --> 00:13:50,000
the results with in the charts. As you399
00:13:50,000 --> 00:13:52,160
can see now the model is connecting to400
00:13:52,160 --> 00:13:55,120
the Google Drive API and uh then search401
00:13:55,120 --> 00:13:57,600
within API it right now it looks like402
00:13:57,600 --> 00:13:59,920
the first result is very relevant. So403
00:13:59,920 --> 00:14:02,720
it's reading the first result.404
00:14:02,720 --> 00:14:04,959
Now it's reading the first result uh in405
00:14:04,959 --> 00:14:07,920
details. Uh let's accelerate this uh406
00:14:07,920 --> 00:14:12,800
replay. So then the model might read407
00:14:12,800 --> 00:14:15,279
from the result again and write some408
00:14:15,279 --> 00:14:16,959
code.409
00:14:16,959 --> 00:14:19,519
So here you can see that the model is410
00:14:19,519 --> 00:14:21,920
using the image generation model called411
00:14:21,920 --> 00:14:24,480
image generation tool to generate some412
00:14:24,480 --> 00:14:28,079
decorations for the slides.413
00:14:28,079 --> 00:14:30,160
And let's see what's the first slide the414
00:14:30,160 --> 00:14:33,399
model made.415
00:14:33,920 --> 00:14:35,920
So here the model is writing some code416
00:14:35,920 --> 00:14:38,399
that will be compiled to be the final417
00:14:38,399 --> 00:14:41,120
slides. So this is the first slide the418
00:14:41,120 --> 00:14:44,160
model make in this demo which looks okay419
00:14:44,160 --> 00:14:46,240
but it's not polished enough.420
00:14:46,240 --> 00:14:48,240
One of the key feature in reinforcement421
00:14:48,240 --> 00:14:50,160
learning is that the model will re422
00:14:50,160 --> 00:14:52,240
review its own results and refine the423
00:14:52,240 --> 00:14:55,120
results to to deliver a good final424
00:14:55,120 --> 00:14:57,839
results. Let's see what's the finally425
00:14:57,839 --> 00:15:00,320
what the model give us.426
00:15:00,320 --> 00:15:04,000
We can click skip and then the model427
00:15:04,000 --> 00:15:07,519
give us a good uh PowerPoint file. So428
00:15:07,519 --> 00:15:09,040
it's a real PowerPoint that you can429
00:15:09,040 --> 00:15:14,040
download and open it in any software.430
00:15:14,639 --> 00:15:19,279
Let's open it in uh in the office. So431
00:15:19,279 --> 00:15:22,160
let's present the slides the model just432
00:15:22,160 --> 00:15:23,839
generated.433
00:15:23,839 --> 00:15:27,120
First are two intelligence benchmarks.434
00:15:27,120 --> 00:15:30,480
Humanities last exam is a benchmark that435
00:15:30,480 --> 00:15:33,519
measures AI's ability to solve a broad436
00:15:33,519 --> 00:15:37,120
range of subjects on hard problems. We437
00:15:37,120 --> 00:15:40,320
evaluate the models with two settings438
00:15:40,320 --> 00:15:43,440
with and without tool use.439
00:15:43,440 --> 00:15:45,920
We can see that the agent modes the raw440
00:15:45,920 --> 00:15:48,720
intelligence is already pretty nice and441
00:15:48,720 --> 00:15:50,880
with access to all tools nearly double442
00:15:50,880 --> 00:15:54,720
the performance to 42%.443
00:15:54,720 --> 00:15:56,720
When evaluating models on humanity's444
00:15:56,720 --> 00:15:59,360
last exam, especially with the browsing445
00:15:59,360 --> 00:16:01,759
ability, we have a two-layer446
00:16:01,759 --> 00:16:04,399
decontamination that ensure that the447
00:16:04,399 --> 00:16:07,680
model doesn't cheat on this benchmark.448
00:16:07,680 --> 00:16:10,079
Front TMS is a benchmark that measures449
00:16:10,079 --> 00:16:11,839
advanced mathematical reasoning ability450
00:16:11,839 --> 00:16:13,680
of models.451
00:16:13,680 --> 00:16:16,000
Different from our baseline of mini and452
00:16:16,000 --> 00:16:18,560
03 which use Python with function453
00:16:18,560 --> 00:16:21,440
coding. We give the agent model all454
00:16:21,440 --> 00:16:23,440
available tools like a browser, a455
00:16:23,440 --> 00:16:26,320
computer and a terminal. The agent456
00:16:26,320 --> 00:16:29,360
achieves new state art of 27% on this457
00:16:29,360 --> 00:16:31,440
benchmark with the help of all these458
00:16:31,440 --> 00:16:34,440
tools.459
00:16:34,639 --> 00:16:36,880
Next, we evaluated the model on two460
00:16:36,880 --> 00:16:39,519
agentic benchmarks. Web arena is a461
00:16:39,519 --> 00:16:41,519
benchmark that measures web agents462
00:16:41,519 --> 00:16:43,600
ability so to solve real world web463
00:16:43,600 --> 00:16:47,279
tasks. The agent model improves over464
00:16:47,279 --> 00:16:51,360
previous O3 model that powers the core.465
00:16:51,360 --> 00:16:54,399
Browse comp is a benchmark we introduced466
00:16:54,399 --> 00:16:56,240
earlier this year that measures the467
00:16:56,240 --> 00:16:58,880
browsing agents ability to search and468
00:16:58,880 --> 00:17:02,320
find uh how to locate information.469
00:17:02,320 --> 00:17:03,839
The agent model significantly470
00:17:03,839 --> 00:17:06,160
outperforms 03 and deep research on this471
00:17:06,160 --> 00:17:11,679
benchmark achieving 69% pass rate.472
00:17:11,679 --> 00:17:14,559
Finally, we care about how the users473
00:17:14,559 --> 00:17:16,959
will benefit from our model in the real474
00:17:16,959 --> 00:17:19,919
world. Spreadsheet bench is a benchmark475
00:17:19,919 --> 00:17:21,919
that measures the model's ability to476
00:17:21,919 --> 00:17:24,400
edit spreadsheets derived from the real477
00:17:24,400 --> 00:17:28,079
world use case. Here the agent model478
00:17:28,079 --> 00:17:30,480
with the liberal office and the computer479
00:17:30,480 --> 00:17:34,000
tool can already solve 30% of the task480
00:17:34,000 --> 00:17:36,480
when we give the model the access to the481
00:17:36,480 --> 00:17:39,840
raw Excel file in the terminal which482
00:17:39,840 --> 00:17:44,000
further boost the performance to 45%.483
00:17:44,000 --> 00:17:46,000
Finally we evated the model on an484
00:17:46,000 --> 00:17:48,000
internal banking benchmark. The bench485
00:17:48,000 --> 00:17:49,760
this benchmark evaluated the model's486
00:17:49,760 --> 00:17:52,559
ability to to conduct first to third487
00:17:52,559 --> 00:17:55,679
year investment bank uh banking analyst488
00:17:55,679 --> 00:17:58,799
tasks such as like putting together a489
00:17:58,799 --> 00:18:00,559
three statement financial model for490
00:18:00,559 --> 00:18:04,000
Fortune uh 500 company in this491
00:18:04,000 --> 00:18:06,160
benchmark. The agent model significantly492
00:18:06,160 --> 00:18:08,080
outperforms the previous deep research493
00:18:08,080 --> 00:18:11,760
and all three models. As you can see494
00:18:11,760 --> 00:18:13,919
this model is one of the most powerful495
00:18:13,919 --> 00:18:16,080
model we've ever trained.496
00:18:16,080 --> 00:18:18,960
It's not only good on benchmarks, it's497
00:18:18,960 --> 00:18:22,480
also capable of reasoning, browsing, and498
00:18:22,480 --> 00:18:24,720
tackling real world tasks at a level499
00:18:24,720 --> 00:18:28,480
that we cannot imagine three months ago.500
00:18:28,480 --> 00:18:31,600
That's right. Um, as Edward said, um, we501
00:18:31,600 --> 00:18:32,799
think we've trained a very powerful502
00:18:32,799 --> 00:18:35,280
model and a lot of the power comes from503
00:18:35,280 --> 00:18:38,240
its ability to browse the internet. And504
00:18:38,240 --> 00:18:40,240
as we know, the internet can be a scary505
00:18:40,240 --> 00:18:42,400
place. There are all sorts of hackers506
00:18:42,400 --> 00:18:45,120
trying to steal your information, scams,507
00:18:45,120 --> 00:18:48,480
uh fishing attempts. Um and agent isn't508
00:18:48,480 --> 00:18:51,120
immune to all these things. Um one509
00:18:51,120 --> 00:18:53,360
particular thing we're worried about is510
00:18:53,360 --> 00:18:55,520
a new uh attack called prompt511
00:18:55,520 --> 00:18:57,120
injections.512
00:18:57,120 --> 00:18:59,840
This is where let's say you ask agent to513
00:18:59,840 --> 00:19:02,080
buy you a book and you give it your514
00:19:02,080 --> 00:19:04,400
credit card information to do that.515
00:19:04,400 --> 00:19:06,240
Agent might stumble upon a malicious516
00:19:06,240 --> 00:19:08,559
website that asks it, "Oh, enter your517
00:19:08,559 --> 00:19:10,400
credit card information here. it'll help518
00:19:10,400 --> 00:19:12,799
you with your task. An agent, which is519
00:19:12,799 --> 00:19:15,200
trained to be helpful, might decide520
00:19:15,200 --> 00:19:18,080
that's a good idea.521
00:19:18,080 --> 00:19:19,760
We've done a lot of work to try to522
00:19:19,760 --> 00:19:22,320
ensure that this doesn't happen. We've523
00:19:22,320 --> 00:19:24,240
trained our model to ignore suspicious524
00:19:24,240 --> 00:19:27,120
instructions on on suspicious websites.525
00:19:27,120 --> 00:19:29,039
We've also have uh we also have layers526
00:19:29,039 --> 00:19:32,000
of monitors that kind of peer over the527
00:19:32,000 --> 00:19:33,760
agent's shoulder and watch it as it's528
00:19:33,760 --> 00:19:36,480
going um and stop the trajectory if529
00:19:36,480 --> 00:19:38,799
anything looks suspicious. We can even530
00:19:38,799 --> 00:19:41,919
update these in real time if new attacks531
00:19:41,919 --> 00:19:44,160
are found in the wild.532
00:19:44,160 --> 00:19:45,919
That said though, you know, this is a533
00:19:45,919 --> 00:19:47,760
cutting edge product. This is a new534
00:19:47,760 --> 00:19:50,000
surface and we can't stop everything.535
00:19:50,000 --> 00:19:51,280
And so that's why I feel it's very536
00:19:51,280 --> 00:19:52,559
important for the audience to be aware537
00:19:52,559 --> 00:19:55,360
of the risks involved in using agent.538
00:19:55,360 --> 00:19:57,440
And um we encourage users to be539
00:19:57,440 --> 00:19:59,520
proactive in kind of thinking about how540
00:19:59,520 --> 00:20:01,120
they share their information. You know,541
00:20:01,120 --> 00:20:02,880
if it's highly sensitive information,542
00:20:02,880 --> 00:20:06,799
maybe don't share that. um maybe um uh543
00:20:06,799 --> 00:20:08,799
use our features like takeover mode to544
00:20:08,799 --> 00:20:10,799
directly input your credit credit card545
00:20:10,799 --> 00:20:12,880
information into the browser instead of546
00:20:12,880 --> 00:20:15,679
um giving it to agent. Um we feel like547
00:20:15,679 --> 00:20:18,640
we've built a very powerful product but548
00:20:18,640 --> 00:20:20,480
again it's important for our users to549
00:20:20,480 --> 00:20:21,760
understand the risk involved.550
00:20:21,760 --> 00:20:23,280
Yeah, I really want to emphasize that I551
00:20:23,280 --> 00:20:25,520
think this is a new level of capability552
00:20:25,520 --> 00:20:27,120
in AI. It's a new way to use AI, but553
00:20:27,120 --> 00:20:28,799
there will be a new set of attacks that554
00:20:28,799 --> 00:20:30,799
come with that. And society and the555
00:20:30,799 --> 00:20:33,120
technology will have to evolve and learn556
00:20:33,120 --> 00:20:34,320
how we're going to mitigate things that557
00:20:34,320 --> 00:20:36,159
we can't even really imagine yet. Uh, as558
00:20:36,159 --> 00:20:37,360
people start doing more and more work559
00:20:37,360 --> 00:20:39,679
this way. Before I wrap up, should we560
00:20:39,679 --> 00:20:41,840
check in on some of the tasks you kicked561
00:20:41,840 --> 00:20:42,080
off?562
00:20:42,080 --> 00:20:46,159
Yeah, let's do it. Um, okay. So, I am563
00:20:46,159 --> 00:20:48,240
going to open a new tab and make sure564
00:20:48,240 --> 00:20:51,840
that we can see the progress of our um,565
00:20:51,840 --> 00:20:55,679
stickers as well. Okay. Let's see. All566
00:20:55,679 --> 00:20:58,159
right. So, sounds like stickers are567
00:20:58,159 --> 00:21:00,880
ready. Let me see what it actually Okay.568
00:21:00,880 --> 00:21:03,200
So, cool thing. This is sort of the end569
00:21:03,200 --> 00:21:06,720
end result of the took about 7 minutes.570
00:21:06,720 --> 00:21:08,480
Highly likely figured out everything.571
00:21:08,480 --> 00:21:09,840
We'll go back and look at the trajectory572
00:21:09,840 --> 00:21:11,679
and see how it did. But at the end573
00:21:11,679 --> 00:21:13,679
result, it looks like it's added to the574
00:21:13,679 --> 00:21:15,360
cart. This is the subtotal. I can just575
00:21:15,360 --> 00:21:17,360
go ahead and look at it and then figure576
00:21:17,360 --> 00:21:20,000
out uh I can just take over at this577
00:21:20,000 --> 00:21:21,600
point as Casey said to enter my credit578
00:21:21,600 --> 00:21:23,039
card information and then place the579
00:21:23,039 --> 00:21:25,200
order really quickly. model is asking580
00:21:25,200 --> 00:21:27,120
for confirmations, etc. as it's supposed581
00:21:27,120 --> 00:21:29,280
to do. Let's just quickly browse through582
00:21:29,280 --> 00:21:31,039
the trajectory and see what it actually583
00:21:31,039 --> 00:21:33,280
did. Oh, it looks like it generated some584
00:21:33,280 --> 00:21:35,840
stickers. Oh, look at that. That's what585
00:21:35,840 --> 00:21:38,880
it generated sticker. Cool. So, yeah,586
00:21:38,880 --> 00:21:40,640
that's the task. I think I can at this587
00:21:40,640 --> 00:21:42,559
point finish up by myself or I can ask588
00:21:42,559 --> 00:21:43,919
the model to actually go ahead and do it589
00:21:43,919 --> 00:21:46,720
for me as well. Let's check on the590
00:21:46,720 --> 00:21:49,840
wedding. Okay, great. Looks like it just591
00:21:49,840 --> 00:21:52,720
finished in the nick of time. Uh, okay,592
00:21:52,720 --> 00:21:55,520
cool. So in this case, as as we said, we593
00:21:55,520 --> 00:21:57,840
were looking for hotel, stress, uh594
00:21:57,840 --> 00:22:01,919
suits, and also shoes. So it's come out595
00:22:01,919 --> 00:22:03,520
with a pretty comprehensive report. It596
00:22:03,520 --> 00:22:05,840
looks like wedding venue, date, when it597
00:22:05,840 --> 00:22:10,240
is with the Zilla links, dress codes. It598
00:22:10,240 --> 00:22:11,600
figured out like what the suit599
00:22:11,600 --> 00:22:12,960
recommendation should be, where you can600
00:22:12,960 --> 00:22:14,799
buy. Now I can go ahead and buy myself601
00:22:14,799 --> 00:22:17,120
or I can ask the agent to go and buy for602
00:22:17,120 --> 00:22:20,960
me. Um also figured out footwear hurdle603
00:22:20,960 --> 00:22:23,360
options. It actually looked through all604
00:22:23,360 --> 00:22:27,120
the oop sorry it looked through all the605
00:22:27,120 --> 00:22:29,360
availability. You can see actually it606
00:22:29,360 --> 00:22:31,440
gives screenshots of what it checked. In607
00:22:31,440 --> 00:22:33,120
this case we use booking.com and it's608
00:22:33,120 --> 00:22:35,280
able to do that. Also has gift609
00:22:35,280 --> 00:22:37,360
suggestions etc. And next step I can ask610
00:22:37,360 --> 00:22:39,760
it as you said the agent says hey if you611
00:22:39,760 --> 00:22:41,520
need assistance purchasing any item or612
00:22:41,520 --> 00:22:42,960
have any further adjustments let me know613
00:22:42,960 --> 00:22:44,880
so we can do that. Um, and I want to614
00:22:44,880 --> 00:22:46,320
show one last demo which we didn't615
00:22:46,320 --> 00:22:48,640
really run live but I think it's really616
00:22:48,640 --> 00:22:51,280
cool and especially because the folks617
00:22:51,280 --> 00:22:52,880
who are getting married are really into618
00:22:52,880 --> 00:22:57,679
MLB. U so we asked the agent uh to go619
00:22:57,679 --> 00:22:59,679
and build an optimal itinary for620
00:22:59,679 --> 00:23:02,640
visiting all 30 MLB stadiums in just in621
00:23:02,640 --> 00:23:05,200
case you're thinking of a satical uh and622
00:23:05,200 --> 00:23:08,159
then design the optimal route prioritize623
00:23:08,159 --> 00:23:10,960
Hello Kitty nights and whatnot and624
00:23:10,960 --> 00:23:12,400
present a final plan as a detailed625
00:23:12,400 --> 00:23:13,520
spreadsheet. I'll really quickly run626
00:23:13,520 --> 00:23:15,440
through this. Um I think it's just so627
00:23:15,440 --> 00:23:18,240
fun to see. So again like as we have628
00:23:18,240 --> 00:23:20,720
thrown shown throughout the the live629
00:23:20,720 --> 00:23:23,919
stream it uses a multitude of tools uses630
00:23:23,919 --> 00:23:26,240
container the terminal use using the631
00:23:26,240 --> 00:23:28,799
browser working through all the details.632
00:23:28,799 --> 00:23:30,400
It'll probably use again back to the633
00:23:30,400 --> 00:23:33,200
browser figuring out Hello Kitty nights634
00:23:33,200 --> 00:23:36,559
and then sports stadium and whatnot. Oh635
00:23:36,559 --> 00:23:39,520
let's see did I miss the Oh go map.636
00:23:39,520 --> 00:23:42,080
building a map using code to actually637
00:23:42,080 --> 00:23:43,919
build it out and then overall we get638
00:23:43,919 --> 00:23:46,159
like a pretty solid result I think at639
00:23:46,159 --> 00:23:48,880
the end takes 25 minutes to work where640
00:23:48,880 --> 00:23:50,400
does the season start and what not you641
00:23:50,400 --> 00:23:51,919
have a spreadsheet that you can quickly642
00:23:51,919 --> 00:23:55,760
view inside just right inside Chad GBD643
00:23:55,760 --> 00:23:57,919
you can map the journey cool looking map644
00:23:57,919 --> 00:24:00,400
I guess and that's it so this is Chad645
00:24:00,400 --> 00:24:02,240
GBD agent we hope you really like it and646
00:24:02,240 --> 00:24:04,000
over to Sam647
00:24:04,000 --> 00:24:05,919
amazing work all of you and and to your648
00:24:05,919 --> 00:24:07,440
teams this is I think uh really649
00:24:07,440 --> 00:24:08,720
something that's going to help people650
00:24:08,720 --> 00:24:10,720
get worked done uh and have more time to651
00:24:10,720 --> 00:24:12,240
do the things they want to do. Um I652
00:24:12,240 --> 00:24:13,520
think it's it's really amazing how much653
00:24:13,520 --> 00:24:15,360
you've brought together to deliver this654
00:24:15,360 --> 00:24:17,760
experience and watching the agent sort655
00:24:17,760 --> 00:24:19,120
of use the internet, make these656
00:24:19,120 --> 00:24:20,640
spreadsheets, make PowerPoints, whatever657
00:24:20,640 --> 00:24:22,960
else uh and do all this work is is quite658
00:24:22,960 --> 00:24:26,000
amazing. We're going live today for pro659
00:24:26,000 --> 00:24:28,880
plus and team users. Pro users will get660
00:24:28,880 --> 00:24:30,720
uh 400 queries a month plus some team661
00:24:30,720 --> 00:24:32,720
users will get 40 a month. Uh the662
00:24:32,720 --> 00:24:34,000
rollout should be finished by the end of663
00:24:34,000 --> 00:24:36,159
the day for pro and very soon for plus664
00:24:36,159 --> 00:24:38,400
and team users. will try to be live for665
00:24:38,400 --> 00:24:40,799
enterprise and edu by the end of this666
00:24:40,799 --> 00:24:43,360
month. As Casey mentioned, although this667
00:24:43,360 --> 00:24:45,360
is an extremely exciting new technology,668
00:24:45,360 --> 00:24:48,080
there are new risks. Uh people learned669
00:24:48,080 --> 00:24:49,520
how to use the internet generally pretty670
00:24:49,520 --> 00:24:50,880
safely, although of course there are671
00:24:50,880 --> 00:24:52,880
still scammers and other attacks. People672
00:24:52,880 --> 00:24:54,559
are going to need to learn to use AI673
00:24:54,559 --> 00:24:56,080
agents. Uh and societyy's going to need674
00:24:56,080 --> 00:24:57,919
to learn to build up defenses against675
00:24:57,919 --> 00:25:00,080
attacks on AI agents as well. So we're676
00:25:00,080 --> 00:25:02,080
starting with a very robust system, lots677
00:25:02,080 --> 00:25:04,240
of warnings. We will relax that over678
00:25:04,240 --> 00:25:05,679
time as people get more comfortable with679
00:25:05,679 --> 00:25:07,600
it. But we do want people to treat this680
00:25:07,600 --> 00:25:09,919
as a new technology and a new risk681
00:25:09,919 --> 00:25:12,080
surface and use all of the caution that682
00:25:12,080 --> 00:25:14,799
Casey talked about. Um, but that said,683
00:25:14,799 --> 00:25:16,720
we hope you'll love it. Uh, this is684
00:25:16,720 --> 00:25:18,159
still very early. We will improve it685
00:25:18,159 --> 00:25:20,640
rapidly and we're excited to see where686
00:25:20,640 --> 00:25:22,640
it all goes. So, congrats again. Thank687
00:25:22,640 --> 00:25:26,440
you very much. Hope you enjoy.