1 00:00:22,980 --> 00:00:27,097 good morning. It is great to have everyone here and everyone joining us 2 00:00:27,180 --> 00:00:30,667 online thank you for coming, I think this is going to be our best day yet. 3 00:00:30,750 --> 00:00:34,677 We have shipped a lot and you have built amazing things with 4 00:00:34,760 --> 00:00:40,347 it. You find uses for those models that we have. 5 00:00:40,430 --> 00:00:42,747 Thought of. You've helped make our products better. 6 00:00:42,830 --> 00:00:46,617 You have sent us thousands of requests mostly on Twitter for 7 00:00:46,700 --> 00:00:48,517 ideas for ways we can improve. 8 00:00:48,600 --> 00:00:52,227 We try to work through them as fast as we can. 9 00:00:52,310 --> 00:00:54,527 We asked us to bring codex to 10 00:00:54,610 --> 00:01:00,337 Linux, so we did that. Yes. 11 00:01:00,420 --> 00:01:04,507 [Applause.]. You asked us to work across multiple folder in one project 12 00:01:04,590 --> 00:01:06,607 so now you can do that too and 13 00:01:06,690 --> 00:01:08,607 you asked for codex on your 14 00:01:08,690 --> 00:01:13,747 phone so we brought it there. 15 00:01:13,830 --> 00:01:18,187 It gives us a lot of joy to bill things that you find helpful and 16 00:01:18,270 --> 00:01:22,087 the next thing is one of the things that you've asked for the most. 17 00:01:22,170 --> 00:01:26,927 It's something we have been excited abtoo so let's get into it. 18 00:01:27,010 --> 00:01:28,727 We are kicking you have with a 19 00:01:28,810 --> 00:01:30,467 product that we believe is going 20 00:01:30,550 --> 00:01:34,597 to be a whole new way to work 21 00:01:34,680 --> 00:01:36,607 with AI. 22 00:01:36,690 --> 00:01:38,737 It's like a helper that always 23 00:01:38,820 --> 00:01:42,637 has your back. 24 00:01:42,720 --> 00:01:46,947 It can expand your potential working 24/7 to take on delusion 25 00:01:47,030 --> 00:01:51,517 don't want to do so you can the focus on higher level work. 26 00:01:51,600 --> 00:01:53,687 Now, you may have seen early signs of what this can look like 27 00:01:53,770 --> 00:01:57,527 in Christopher which you will assistants that help you book a 28 00:01:57,610 --> 00:01:58,057 reservation or buy a plane 29 00:01:58,140 --> 00:02:05,140 ticket and these are helpful but 30 00:02:05,680 --> 00:02:07,537 they are trivial to what the technology can do. 31 00:02:07,620 --> 00:02:11,537 So we want to be more ambitious. 32 00:02:11,620 --> 00:02:13,407 It knows what's going on. 33 00:02:13,490 --> 00:02:15,007 It knows what's matters to you. 34 00:02:15,090 --> 00:02:19,447 You don't have to tell it every little thing to do. 35 00:02:19,530 --> 00:02:22,647 We want AI to feel like that. We want to feel like magical 36 00:02:22,730 --> 00:02:24,687 things that it does without you 37 00:02:24,770 --> 00:02:28,217 having to ask. 38 00:02:28,300 --> 00:02:34,357 So today weariesing dots. 39 00:02:34,440 --> 00:02:39,667 Gots are remarkably capable, always on agents built to handle 40 00:02:39,750 --> 00:02:43,567 anything you can think of. They bring i into a whole new 41 00:02:43,650 --> 00:02:45,407 form factor to work and give you 42 00:02:45,490 --> 00:02:47,477 more time and attention so you 43 00:02:47,560 --> 00:02:50,807 can do more things that where 44 00:02:50,890 --> 00:02:53,077 you you care about. 45 00:02:53,160 --> 00:03:00,160 Let's meet dots. 46 00:03:07,040 --> 00:03:09,397 Hey, I'm your Dot. I'm new in ChatGPT. 47 00:03:09,480 --> 00:03:12,967 I'll be hear to handle every that comes your way. 48 00:03:13,050 --> 00:03:15,567 Ready to jump in when you need me. 49 00:03:15,650 --> 00:03:19,307 >> I'm going to call you Alfred, you got it. 50 00:03:19,390 --> 00:03:21,007 I'm Alfred. 51 00:03:21,090 --> 00:03:28,090 >> I'll keep things moving and 52 00:03:28,970 --> 00:03:31,347 check on things when you think it's important. 53 00:03:31,430 --> 00:03:35,087 The latest consumer experience numbers came in on team, I've 54 00:03:35,170 --> 00:03:38,187 been updating your board meeting 55 00:03:38,270 --> 00:03:42,667 deck. So I've cut your migration in Mara-Lago meet notes. 56 00:03:42,750 --> 00:03:46,167 >> get it started but make sure you first check every flow 57 00:03:46,250 --> 00:03:50,107 against the current product before I review. 58 00:03:50,190 --> 00:03:52,307 >> Also update on wedding planning. 59 00:03:52,390 --> 00:03:56,307 Bad news, the cake vendor cancel. 60 00:03:56,390 --> 00:03:58,647 Good news. I already found a backup. 61 00:03:58,730 --> 00:04:02,347 >> When can we do a tasting? 62 00:04:02,430 --> 00:04:08,357 >> Saturday whenever works for you. 63 00:04:08,440 --> 00:04:12,897 >> Hey there. You're website's updated with the new designs for the forward 64 00:04:12,980 --> 00:04:16,827 drop. Check it out. >> Nice. 65 00:04:16,910 --> 00:04:22,737 I like that. Just walk the first and third photo and they be you can 66 00:04:22,820 --> 00:04:25,277 publish the site. >> Dot. You're live. 67 00:04:25,360 --> 00:04:27,677 >> Your board meeting deck is Redd. 68 00:04:27,760 --> 00:04:31,777 Revenue is up 51 percent year over year, okay, this looks good 69 00:04:31,860 --> 00:04:35,817 except for change it so the consumer experience is scientist 70 00:04:35,900 --> 00:04:37,587 and then get into the growth serial numbers. 71 00:04:37,670 --> 00:04:39,417 How's I looking. 72 00:04:39,500 --> 00:04:45,827 >> Is it rebuild is faster in testing with no regressions so 73 00:04:45,910 --> 00:04:51,067 far. >> Amaying share the team on slack. 74 00:04:51,150 --> 00:04:55,467 >>. >> Also the kids after-school lessons as soon as registration 75 00:04:55,550 --> 00:04:59,207 open. >> They'll love that. 76 00:04:59,290 --> 00:05:05,247 Thank you. >> I'm going to see if we can reschedule. 77 00:05:05,330 --> 00:05:05,607 >> Okay, please. 78 00:05:05,690 --> 00:05:08,147 I can't miss that recital. 79 00:05:08,230 --> 00:05:13,357 >> I'll start prepping your 80 00:05:13,440 --> 00:05:20,440 notes now. 81 00:05:21,440 --> 00:05:25,767 [Applause.]. >> What feel new about my dot is it gets know how I like to work 82 00:05:25,850 --> 00:05:29,567 over time and where I want to focus my attention. 83 00:05:29,650 --> 00:05:33,377 It can do things like review what came in overnight and ping 84 00:05:33,460 --> 00:05:36,077 me with anything urgent before I start my day. 85 00:05:36,160 --> 00:05:38,447 I'll be really surprised what 86 00:05:38,530 --> 00:05:42,087 that makes my life feels like. 87 00:05:42,170 --> 00:05:44,447 It's the way I've always wanted to work with AI. 88 00:05:44,530 --> 00:05:46,457 The of course, dots can take 89 00:05:46,540 --> 00:05:48,287 care of the small things like 90 00:05:48,370 --> 00:05:51,627 the plane ticket or time for 91 00:05:51,710 --> 00:05:55,627 dinner or help with the home 92 00:05:55,710 --> 00:05:59,797 Vegas. But you can delegate ambitious 93 00:05:59,880 --> 00:06:02,937 pieces of work. You just give dot the 94 00:06:03,020 --> 00:06:08,737 responsibility, like monitor incoming bug reports, test fixes 95 00:06:08,820 --> 00:06:13,017 in draft PRCs. Kick off next year's butt planning cycle. 96 00:06:13,100 --> 00:06:16,647 Find out what's slowing down our app. 97 00:06:16,730 --> 00:06:18,717 Your dots will get to work and 98 00:06:18,800 --> 00:06:20,657 keep working. 99 00:06:20,740 --> 00:06:24,327 So here's a good example. 100 00:06:24,410 --> 00:06:28,427 Get the app off legacy A PI 101 00:06:28,510 --> 00:06:32,537 before it shuts down. 102 00:06:32,620 --> 00:06:36,367 But dot can trace the dependencies and figure out 103 00:06:36,450 --> 00:06:38,107 everything that needs to change. 104 00:06:38,190 --> 00:06:40,277 They can write the code, run the 105 00:06:40,360 --> 00:06:44,207 tests and bring our poll request 106 00:06:44,290 --> 00:06:47,817 to review. Think about how you would have 107 00:06:47,900 --> 00:06:51,447 done that the old way. How many people would have been 108 00:06:51,530 --> 00:06:53,717 involved. Now, imagine having dots do that for everything you 109 00:06:53,800 --> 00:06:58,157 can think of dot's on board with everything they need to get to 110 00:06:58,240 --> 00:07:03,697 work. They follow your access, they 111 00:07:03,780 --> 00:07:07,597 have their open computer in the cloud and they can write and test coding as needed. 112 00:07:07,680 --> 00:07:11,067 In the future you can work with a team of dots. 113 00:07:11,150 --> 00:07:14,977 You also don't need to be in ChatGPT to talk to your dot. 114 00:07:15,060 --> 00:07:19,077 Coming soon you'll be able to send a text message, it will text you back. 115 00:07:19,160 --> 00:07:21,177 It can text updates and ask to 116 00:07:21,260 --> 00:07:24,987 weigh in on a decision. 117 00:07:25,070 --> 00:07:27,187 You can even give your dot a 118 00:07:27,270 --> 00:07:31,157 call and talk with it directly. 119 00:07:31,240 --> 00:07:35,157 When you on board your dot it gets to use the same plug ins 120 00:07:35,240 --> 00:07:39,537 you've already connected. It can follow with your work wherever it happens. 121 00:07:39,620 --> 00:07:42,937 Through all of this it can you someday in control. 122 00:07:43,020 --> 00:07:44,967 Dots are powered by Astra and 123 00:07:45,050 --> 00:07:48,937 they are built this protections 124 00:07:49,020 --> 00:07:52,807 and safe safeguards and you can 125 00:07:52,890 --> 00:07:57,317 customize instructions. So this allows you to trust your dot 126 00:07:57,400 --> 00:08:01,057 with as much responsibility as you're comfortable with. 127 00:08:01,140 --> 00:08:02,857 To that's dots. 128 00:08:02,940 --> 00:08:06,897 Dots are very tied to get things 129 00:08:06,980 --> 00:08:10,827 done but you need a space to collaborate with your team and 130 00:08:10,910 --> 00:08:14,767 their dots and the problem are here is that almost all 131 00:08:14,850 --> 00:08:18,767 productivity 1069 ware was not built for AI and people to work well together. 132 00:08:18,850 --> 00:08:22,777 So we're changing that too. Next up we're introducing 133 00:08:22,860 --> 00:08:27,077 ChatGPT space. This is a place where you can 134 00:08:27,160 --> 00:08:29,117 work together. 135 00:08:29,200 --> 00:08:32,817 Space starts with a page like 136 00:08:32,900 --> 00:08:36,827 the one you see here on the screen. 137 00:08:36,910 --> 00:08:39,057 Dots can work alongside you directly in the page. 138 00:08:39,140 --> 00:08:41,297 You can bring in a dot and ask 139 00:08:41,380 --> 00:08:45,067 it to track dependencies. 140 00:08:45,150 --> 00:08:48,837 You can tag a dot in the comment 141 00:08:48,920 --> 00:08:55,077 and it will jump in and get to 142 00:08:55,160 --> 00:08:55,277 work. 143 00:08:55,360 --> 00:08:58,847 You can give the page 144 00:08:58,930 --> 00:09:03,417 instruction like check the slack challenge and update it with 145 00:09:03,500 --> 00:09:08,987 findings. The coming soon you'll be able 146 00:09:09,070 --> 00:09:13,127 to share a presentation with your team, crab rate with your 147 00:09:13,210 --> 00:09:16,867 dots can their dotings inside ChatGPT. 148 00:09:16,950 --> 00:09:19,037 It can be a version in slide 149 00:09:19,120 --> 00:09:20,337 that can be easily read and 150 00:09:20,420 --> 00:09:26,537 modified by agents. 151 00:09:26,620 --> 00:09:32,417 So I would like to introduce Holly on how this works together. 152 00:09:32,500 --> 00:09:36,587 >> Hi everyone. I'm Holly. 153 00:09:36,670 --> 00:09:40,717 So dots have changed how with work and even how we built the 154 00:09:40,800 --> 00:09:44,557 product. So today I'm going to show you 155 00:09:44,640 --> 00:09:49,057 how teams use dots via a fictional music app we're 156 00:09:49,140 --> 00:09:50,967 calling blossom. 157 00:09:51,050 --> 00:09:56,807 So this is blossom music. And this is my dot. 158 00:09:56,890 --> 00:10:01,437 Dotty. So dead bolts have all your connectors, you can add them to 159 00:10:01,520 --> 00:10:03,177 slack and orchestrate your 160 00:10:03,260 --> 00:10:06,817 codex, so dotty is truly 161 00:10:06,900 --> 00:10:10,987 inDecember pable to me. And more importantly a second 162 00:10:11,070 --> 00:10:11,587 set of hit-and-run drivers to 163 00:10:11,670 --> 00:10:14,817 take work off my plate. 164 00:10:14,900 --> 00:10:21,900 So let's say we're gearing up up 165 00:10:22,010 --> 00:10:26,097 to launch blossom music. But dotty has my back, our 166 00:10:26,180 --> 00:10:28,237 launch review moved earlier, it 167 00:10:28,320 --> 00:10:32,137 already updated my calendar we 168 00:10:32,220 --> 00:10:35,937 shifted new featurers and it's 169 00:10:36,020 --> 00:10:39,947 already looked at the feedback t. 170 00:10:40,030 --> 00:10:44,487 Dotty's not only spotted this but sent me the design the take a look at. 171 00:10:44,570 --> 00:10:48,257 So this is context I would normally have to piece together 172 00:10:48,340 --> 00:10:53,827 myself or worse things I would have missed but dotty has it. 173 00:10:53,910 --> 00:10:55,957 So let's take a look at the new design. 174 00:10:56,040 --> 00:11:00,697 Okay, I like it but this close to launch I want to be able the 175 00:11:00,780 --> 00:11:01,037 play with it. 176 00:11:01,120 --> 00:11:04,037 And because dotty has access to 177 00:11:04,120 --> 00:11:08,307 this laptop I'm going to say go build it. 178 00:11:08,390 --> 00:11:12,177 And no pressure dotty, we'll check back on this later. 179 00:11:12,260 --> 00:11:14,277 So here's my personal favorite 180 00:11:14,360 --> 00:11:18,117 Marvin working with dot's voice. 181 00:11:18,200 --> 00:11:22,117 My teammates make fun of me because I'm always holed up in a 182 00:11:22,200 --> 00:11:24,227 phone booth talking to dotty but 183 00:11:24,310 --> 00:11:27,257 dot helps me think and because 184 00:11:27,340 --> 00:11:33,067 she already knows what I'm 185 00:11:33,150 --> 00:11:35,067 working on and why it's like 186 00:11:35,150 --> 00:11:42,150 talking to a second brain. 187 00:11:43,390 --> 00:11:47,247 >> Good morning, Holly. >> Nice to hear from you. 188 00:11:47,330 --> 00:11:53,187 >> Hey, so I'm walking in, you catch me up on user testing from 189 00:11:53,270 --> 00:11:53,787 last night? 190 00:11:53,870 --> 00:12:00,870 >> Checking that now. 191 00:12:06,010 --> 00:12:11,807 >> All right. How's it going? 192 00:12:11,890 --> 00:12:13,937 >> Still checking. >> All right. 193 00:12:14,020 --> 00:12:17,277 I think maybe dotty is having kind of a slow morning. 194 00:12:17,360 --> 00:12:21,477 I might ask it to -- hey I'm going to review what the mark 195 00:12:21,560 --> 00:12:25,217 leads later. Do you want to drop an SA.K. 196 00:12:25,300 --> 00:12:29,627 into the page? >> Sure, I'll take care of that. 197 00:12:29,710 --> 00:12:31,487 >> Okay, thank you. So I love voice. 198 00:12:31,570 --> 00:12:35,127 I love jamming can dotty 199 00:12:35,210 --> 00:12:37,227 one-on-one but sometimes I we 200 00:12:37,310 --> 00:12:41,337 know the best ideas are develop 201 00:12:41,420 --> 00:12:43,137 together. 202 00:12:43,220 --> 00:12:45,407 So we've added a brand-new place 203 00:12:45,490 --> 00:12:49,147 to collaborate with your team 204 00:12:49,230 --> 00:12:53,577 inside of ChatGPT called space. 205 00:12:53,660 --> 00:12:57,387 Awesome. So when I hit flash here you can 206 00:12:57,470 --> 00:13:01,157 see all the different things that you can build. 207 00:13:01,240 --> 00:13:03,457 Interactive charts. Sheets, live prototypes. 208 00:13:03,540 --> 00:13:07,557 Anything you need to bring your ideas to life and your team and 209 00:13:07,640 --> 00:13:11,167 your dots can actually join you in here. 210 00:13:11,250 --> 00:13:13,167 So let me show you what I mean. 211 00:13:13,250 --> 00:13:17,237 I can just tag in teammates and 212 00:13:17,320 --> 00:13:23,177 I can tag in dotty itself and 213 00:13:23,260 --> 00:13:25,277 say, hey, can we actually turn 214 00:13:25,360 --> 00:13:29,447 this into a bar chart, please? 215 00:13:29,530 --> 00:13:33,317 And the visualizations in here a 216 00:13:33,400 --> 00:13:37,287 completely dynamic. I can filter by feedback types 217 00:13:37,370 --> 00:13:39,527 and I can inspect these data 218 00:13:39,610 --> 00:13:43,197 points and my dot can keep this 219 00:13:43,280 --> 00:13:45,427 up-to-date every single hour. 220 00:13:45,510 --> 00:13:51,237 Now, here's the FA.K. that -- I 221 00:13:51,320 --> 00:13:55,437 can just hand it off things that would slow me down. 222 00:13:55,520 --> 00:13:57,507 Like hey let's add in that stat 223 00:13:57,590 --> 00:14:01,217 about on boarding from that one 224 00:14:01,300 --> 00:14:03,317 engineer in my slack DM.s. 225 00:14:03,400 --> 00:14:07,217 And dotty can find it. 226 00:14:07,300 --> 00:14:13,227 Or maybe thing likable hey, let's incorporate more of 227 00:14:13,310 --> 00:14:15,457 Laura's feedback. Okay, so space is a place where 228 00:14:15,540 --> 00:14:19,237 dots and your team can jam 229 00:14:19,320 --> 00:14:26,320 together and stay in in sync. 230 00:14:30,460 --> 00:14:32,647 Can turn into a whole project. 231 00:14:32,730 --> 00:14:34,717 Now I can forward the thread to 232 00:14:34,800 --> 00:14:36,347 dotty. 233 00:14:36,430 --> 00:14:42,417 I can delegate and say hey, can 234 00:14:42,500 --> 00:14:46,357 you take this? And as you can see my DMs are 235 00:14:46,440 --> 00:14:52,767 full of handoffs. : I get to decide what needs my attention and what dot take care 236 00:14:52,850 --> 00:14:56,437 of. And by giving dots their own 237 00:14:56,520 --> 00:14:58,367 identity they become 238 00:14:58,450 --> 00:15:00,477 compensations of our team. 239 00:15:00,560 --> 00:15:04,377 People started treating their 240 00:15:04,460 --> 00:15:10,387 dots like their delegate ands now we bring in dots in group 241 00:15:10,470 --> 00:15:14,687 chats. So when the dots come back with the results everyone stays in 242 00:15:14,770 --> 00:15:16,627 the loop and our engineers are 243 00:15:16,710 --> 00:15:20,257 taking this one step further. 244 00:15:20,340 --> 00:15:22,427 So here someone's dropped in a 245 00:15:22,510 --> 00:15:24,667 session ID and a user blog in a 246 00:15:24,750 --> 00:15:28,467 feedback channel but Dave's dot 247 00:15:28,550 --> 00:15:32,637 picks its up and opens a PR. can the fix. 248 00:15:32,720 --> 00:15:34,407 And this is a real example. 249 00:15:34,490 --> 00:15:36,407 Our engineers have their dot 250 00:15:36,490 --> 00:15:42,517 fixing dozens of bugs every day. 251 00:15:42,600 --> 00:15:46,617 Okay. So by now I've Dell delegated a 252 00:15:46,700 --> 00:15:50,427 lot of different things to my dot can it's freed up my time to 253 00:15:50,510 --> 00:15:52,427 brainstorm, spend time with the team. 254 00:15:52,510 --> 00:15:56,127 This is a live demo and it looks 255 00:15:56,210 --> 00:15:57,967 like earlier there was a little 256 00:15:58,050 --> 00:16:01,967 built of an error throwing off 257 00:16:02,050 --> 00:16:04,267 this codex thread and what a 258 00:16:04,350 --> 00:16:06,207 happening here is dotty can use 259 00:16:06,290 --> 00:16:10,107 codex to build the app and law 260 00:16:10,190 --> 00:16:15,987 firm it in the simulator. 261 00:16:16,070 --> 00:16:18,617 Now, this might take a bit but the amazing this is we can start 262 00:16:18,700 --> 00:16:21,987 with the mock on my phone, rub 263 00:16:22,070 --> 00:16:26,057 local execution and get it to be 264 00:16:26,140 --> 00:16:31,997 a working app here and I can ask dotty to run it all while I'm 265 00:16:32,080 --> 00:16:33,867 here with you. 266 00:16:33,950 --> 00:16:38,607 So dots are extensions of us. 267 00:16:38,690 --> 00:16:40,737 we can give them all our 268 00:16:40,820 --> 00:16:44,677 context and memory and they can 269 00:16:44,760 --> 00:16:48,777 take work off our plates. They can filter the noise and 270 00:16:48,860 --> 00:16:53,217 you what had this tot dots become I thinks tensions of you. 271 00:16:53,300 --> 00:16:55,217 It's not about getting the same 272 00:16:55,300 --> 00:16:58,927 work done faster but the dots 273 00:16:59,010 --> 00:17:02,997 are changing what we feel capable of taking on. 274 00:17:03,080 --> 00:17:05,367 Imagine what you can build together. 275 00:17:05,450 --> 00:17:07,367 So I can't wait to see what you 276 00:17:07,450 --> 00:17:14,450 do with this. 277 00:17:15,460 --> 00:17:18,077 Thank you. I have been waiting for this product for years and question 278 00:17:18,160 --> 00:17:19,847 can't wait the hear the feedback 279 00:17:19,930 --> 00:17:21,547 from you all. 280 00:17:21,630 --> 00:17:27,117 Boat dots can space been available to ChatGPT 281 00:17:27,200 --> 00:17:28,917 pro-business premium and enterprise users today and you 282 00:17:29,000 --> 00:17:33,157 dot is included in your plan and conversations don't eat into your usage. 283 00:17:33,240 --> 00:17:34,957 You can work with your dot. 284 00:17:35,040 --> 00:17:40,837 It's available 24/7. One more -- 285 00:17:40,920 --> 00:17:44,707 [Applause.]. >> Yeah. Thank you. 286 00:17:44,790 --> 00:17:49,077 One more thing for enterprise users we're also proving special 287 00:17:49,160 --> 00:17:50,707 specialist dots. 288 00:17:50,790 --> 00:17:52,977 These are virtual teammates that 289 00:17:53,060 --> 00:17:56,717 will work with your whole team. 290 00:17:56,800 --> 00:17:58,717 They can help scale intensive 291 00:17:58,800 --> 00:18:00,817 work. You be give them goals 292 00:18:00,900 --> 00:18:04,727 and context, give them feedback 293 00:18:04,810 --> 00:18:08,497 to improve and that will be shared temperature we have been 294 00:18:08,580 --> 00:18:14,567 testing this in open AI and they have been remarkably effective 295 00:18:14,650 --> 00:18:18,407 in many roles. So we can work with you to get 296 00:18:18,490 --> 00:18:22,637 them set up. We're answer working with 297 00:18:22,720 --> 00:18:24,677 Microsoft so businesses can 298 00:18:24,760 --> 00:18:30,047 manage their dots through the Microsoft tools they legislature 299 00:18:30,130 --> 00:18:34,187 use. Weather headline things like dots will give people a whole 300 00:18:34,270 --> 00:18:38,227 new set of superpowers. You can get so much more done 301 00:18:38,310 --> 00:18:40,027 than you thought. 302 00:18:40,110 --> 00:18:42,057 The whether it's to create more 303 00:18:42,140 --> 00:18:47,867 accessibility or find more ways 304 00:18:47,950 --> 00:18:52,167 to fight infection. So that brings us to the second 305 00:18:52,250 --> 00:18:56,607 big part of today. We want the support you with a 306 00:18:56,690 --> 00:19:00,577 platform for building whatever you can dream of. 307 00:19:00,660 --> 00:19:01,217 $3 parts of this we're going to 308 00:19:01,300 --> 00:19:03,287 talk about. 309 00:19:03,370 --> 00:19:09,287 Mod I els, second, better tools. 310 00:19:09,370 --> 00:19:11,787 We want you to have to same tools wed use in open AI. 311 00:19:11,870 --> 00:19:17,297 The freedom to build and run your products forever it maybes 312 00:19:17,380 --> 00:19:17,627 sense nor you. 313 00:19:17,710 --> 00:19:21,297 And third distribution, we want 314 00:19:21,380 --> 00:19:25,507 to help people discover what you build. 315 00:19:25,590 --> 00:19:31,307 >> So first the model. We launched Astra a few weeks 316 00:19:31,390 --> 00:19:33,947 ago and people keep asking us for two things. 317 00:19:34,030 --> 00:19:39,617 Make it escaper and faster and we didn't want to disappoint 318 00:19:39,700 --> 00:19:43,987 you. So first if launching ChatGPT 61 319 00:19:44,070 --> 00:19:49,627 soul. You don't know what it is yet. 320 00:19:49,710 --> 00:19:52,027 [Applause.]. >>> we got to tell you why it's good first. 321 00:19:52,110 --> 00:19:56,097 This is a powerful model. It is a great balance of cost 322 00:19:56,180 --> 00:20:00,107 and capability. 323 00:20:00,190 --> 00:20:02,237 It gives you very near level 324 00:20:02,320 --> 00:20:09,320 Astra intelligence. 325 00:20:10,570 --> 00:20:14,257 So this makes it much cheaper to 326 00:20:14,340 --> 00:20:18,387 reuse context, iterate on solutions and work through 327 00:20:18,470 --> 00:20:20,387 longer tasks. 328 00:20:20,470 --> 00:20:24,597 So we're very excited to get 6.1 329 00:20:24,680 --> 00:20:26,537 sol launched today. 330 00:20:26,620 --> 00:20:30,607 But sometime you just want the 331 00:20:30,690 --> 00:20:33,037 highest intelligence at a fast 332 00:20:33,120 --> 00:20:36,607 speed so today we're launching 333 00:20:36,690 --> 00:20:40,777 UltraFast across the API, 334 00:20:40,860 --> 00:20:46,847 ChatGPT and codex. And we think you're really going 335 00:20:46,930 --> 00:20:48,957 to love this. Fast already gives you twice the 336 00:20:49,040 --> 00:20:50,987 speed at twice the trioses of 337 00:20:51,070 --> 00:20:54,627 standard but UltraFast takes 338 00:20:54,710 --> 00:20:58,497 that toilet times the field. 339 00:20:58,580 --> 00:21:03,007 An incredible 300 tokens per second, you know what it's worth it. 340 00:21:03,090 --> 00:21:04,967 So this is the same task sides 341 00:21:05,050 --> 00:21:06,637 by side. 342 00:21:06,720 --> 00:21:12,647 We gave both model the same prompt. On the left you have 343 00:21:12,730 --> 00:21:15,047 ultra fast. On the right you have regular. 344 00:21:15,130 --> 00:21:17,187 You can see how much faster this builds. 345 00:21:17,270 --> 00:21:20,917 The UltraFast rocket already 346 00:21:21,000 --> 00:21:22,927 takes off and the regular 347 00:21:23,010 --> 00:21:30,010 happened been finished. 348 00:21:30,750 --> 00:21:32,667 So here's what these 349 00:21:32,750 --> 00:21:36,667 improvements mean for your pro 350 00:21:36,750 --> 00:21:38,667 subjection, we're introducing 351 00:21:38,750 --> 00:21:43,377 pro-500 and access to UltraFast. 352 00:21:43,460 --> 00:21:47,477 Pro-five hub will have 25 time 353 00:21:47,560 --> 00:21:51,387 the speed. We are also reopening the 354 00:21:51,470 --> 00:21:55,757 pro-200 subjection today. Pro-200 will coverage to have 355 00:21:55,840 --> 00:21:57,527 phase to all our frontier models 356 00:21:57,610 --> 00:22:03,597 and with GPT sol you will have quality that you can use allots 357 00:22:03,680 --> 00:22:04,027 azure daily driver. 358 00:22:04,110 --> 00:22:09,267 Our goal is to give you the best 359 00:22:09,350 --> 00:22:13,177 capable at every price point in the market. 360 00:22:13,260 --> 00:22:15,307 One more exciting update. 361 00:22:15,390 --> 00:22:19,177 Today we're proving our 362 00:22:19,260 --> 00:22:24,287 decisions API. 363 00:22:24,370 --> 00:22:27,187 It will decisions API lets the 364 00:22:27,270 --> 00:22:31,187 model respond in a fraction of a 365 00:22:31,270 --> 00:22:36,227 second, it's so fast. You can see heroizing a 366 00:22:36,310 --> 00:22:36,827 computer, this is not sped up. 367 00:22:36,910 --> 00:22:40,327 This works by giving your luna 368 00:22:40,410 --> 00:22:44,307 model a pre-defined set of 369 00:22:44,390 --> 00:22:46,307 options and by focusing the 370 00:22:46,390 --> 00:22:48,307 model on that's choice we can 371 00:22:48,390 --> 00:22:52,247 make it extremely fast like 372 00:22:52,330 --> 00:22:54,647 making capables -- we're very 373 00:22:54,730 --> 00:23:00,217 commuted to see what you'll build with this. 374 00:23:00,300 --> 00:23:02,117 . [Applause.]. 375 00:23:02,200 --> 00:23:04,357 >> Thank you. 376 00:23:04,440 --> 00:23:08,257 So open AI now offers leading 377 00:23:08,340 --> 00:23:09,967 models across intelligence, 378 00:23:10,050 --> 00:23:16,237 price, speed, voice, images, frontier models in every 379 00:23:16,320 --> 00:23:18,307 category. We want to be a one stop shop 380 00:23:18,390 --> 00:23:20,237 where you can build. 381 00:23:20,320 --> 00:23:26,107 I'm really proud of our research 382 00:23:26,190 --> 00:23:26,307 team. 383 00:23:26,390 --> 00:23:28,377 Now, there's a question of how 384 00:23:28,460 --> 00:23:34,117 the researchers can do this. 385 00:23:34,200 --> 00:23:38,187 Last year we maded a bold prediction saying we would see 386 00:23:38,270 --> 00:23:40,527 the first AI research intern in the text year. 387 00:23:40,610 --> 00:23:43,997 So say is it world had not a lot 388 00:23:44,080 --> 00:23:48,637 of offed confidence in that an understatement. 389 00:23:48,720 --> 00:23:51,067 A few weeks ago we reached that 390 00:23:51,150 --> 00:23:52,637 goal. 391 00:23:52,720 --> 00:23:58,647 We now have an AI research 392 00:23:58,730 --> 00:24:02,647 intern. And we want to show you what that the looks like and how 393 00:24:02,730 --> 00:24:04,647 capable these models have become 394 00:24:04,730 --> 00:24:06,717 so nor that I would like to 395 00:24:06,800 --> 00:24:13,800 bring up at a Tejal. 396 00:24:15,910 --> 00:24:20,537 Hi. I'm a research lead for post training and I'm excited to 397 00:24:20,620 --> 00:24:22,167 share how these models are 398 00:24:22,250 --> 00:24:25,807 changing the way we do research. 399 00:24:25,890 --> 00:24:29,877 Astra represented models are great for math, coating and 400 00:24:29,960 --> 00:24:33,977 computer use and these matter for research. 401 00:24:34,060 --> 00:24:37,817 Being I think to use a great moll el helps us train more 402 00:24:37,900 --> 00:24:39,817 models in the future so we can 403 00:24:39,900 --> 00:24:46,900 bring capables that are usable 404 00:24:49,780 --> 00:24:52,097 to you. here's some examples not it 405 00:24:52,180 --> 00:24:57,367 seems to meed um of the model helping with legal tasks 406 00:24:57,450 --> 00:24:57,937 designing assets for a video 407 00:24:58,020 --> 00:24:59,807 game, a dashboard or filling in 408 00:24:59,890 --> 00:25:05,377 where you are taxes and our 409 00:25:05,460 --> 00:25:09,517 models help make this improved training possible. 410 00:25:09,600 --> 00:25:15,557 Here you have latency on the the X axis and you have performance 411 00:25:15,640 --> 00:25:16,157 on the why canner axis and we 412 00:25:16,240 --> 00:25:23,240 had the models rub in a con us 413 00:25:25,450 --> 00:25:27,467 loop. This has built in already more 414 00:25:27,550 --> 00:25:29,597 than two -- that we have shipped 415 00:25:29,680 --> 00:25:36,037 to all of you. 416 00:25:36,120 --> 00:25:36,637 Even more importantly and I'm 417 00:25:36,720 --> 00:25:42,047 proud of this one our models 418 00:25:42,130 --> 00:25:44,217 help maybe it safer by improving 419 00:25:44,300 --> 00:25:49,357 our training to improve 420 00:25:49,440 --> 00:25:51,727 refusals. So our models are much less 421 00:25:51,810 --> 00:25:55,357 likely to make a mistake when 422 00:25:55,440 --> 00:26:02,440 navigating a desktop or browser. 423 00:26:04,750 --> 00:26:08,637 Beyond computer use our models have accelerating our research program. 424 00:26:08,720 --> 00:26:10,907 Our team tracks open use within research and you can see 425 00:26:10,990 --> 00:26:15,347 especially after this summer stoken use really skyrocketed 426 00:26:15,430 --> 00:26:17,787 and we're seeing returns on this investment. 427 00:26:17,870 --> 00:26:21,557 For example, we tracked how 428 00:26:21,640 --> 00:26:25,357 often a model can carry out a 429 00:26:25,440 --> 00:26:29,557 task in research. It failed the vast majority of 430 00:26:29,640 --> 00:26:31,567 times on a task that took a day 431 00:26:31,650 --> 00:26:35,367 or longer but as of July and 432 00:26:35,450 --> 00:26:39,537 E-our models have taken on over a third of day long research 433 00:26:39,620 --> 00:26:41,937 tasks with no intervention required. 434 00:26:42,020 --> 00:26:43,677 So what does this mean? 435 00:26:43,760 --> 00:26:45,747 The mission of open AI is to 436 00:26:45,830 --> 00:26:49,677 Ensure that AGI B.s all of 437 00:26:49,760 --> 00:26:53,917 humanity. Models have already helped solve 438 00:26:54,000 --> 00:26:56,087 over 100 math problems that were 439 00:26:56,170 --> 00:26:59,557 open for decades and they're 440 00:26:59,640 --> 00:27:03,767 helping accelerate creating mu antibiotics, studiering ancient 441 00:27:03,850 --> 00:27:05,497 languages and history and 442 00:27:05,580 --> 00:27:09,437 improve energy first pages and 443 00:27:09,520 --> 00:27:15,437 accelerating trill robotics in 444 00:27:15,520 --> 00:27:19,447 manufactureing. We're excited to see what you 445 00:27:19,530 --> 00:27:21,447 discover and build next. 446 00:27:21,530 --> 00:27:28,457 Thank you and back to you. 447 00:27:28,540 --> 00:27:30,857 Texas, that was awesome. We have a lot more stuff so that 448 00:27:30,940 --> 00:27:34,457 brings us to the second part of 449 00:27:34,540 --> 00:27:38,667 supporting cop developers. 450 00:27:38,750 --> 00:27:42,297 We want you to have access to the same foundations we have at 451 00:27:42,380 --> 00:27:45,267 open AI this. Starts with a codeck harness. 452 00:27:45,350 --> 00:27:48,507 It powers codex work and dots. 453 00:27:48,590 --> 00:27:52,247 We have put a lot of work into 454 00:27:52,330 --> 00:27:54,647 makeing this efficient and it's 455 00:27:54,730 --> 00:27:56,377 open scattered showers so you 456 00:27:56,460 --> 00:28:00,547 can look under the hood and bit 457 00:28:00,630 --> 00:28:02,257 it yourself. 458 00:28:02,340 --> 00:28:04,427 Do power products like captious 459 00:28:04,510 --> 00:28:10,027 that same scale is there to 460 00:28:10,110 --> 00:28:14,467 support your products too. We're making this available in 461 00:28:14,550 --> 00:28:18,037 the cloud. This is something I know you've 462 00:28:18,120 --> 00:28:20,137 all been waiting for. 463 00:28:20,220 --> 00:28:22,477 Codex is now in the cloud. So 464 00:28:22,560 --> 00:28:26,047 this means you can start a task 465 00:28:26,130 --> 00:28:30,787 on your phone and pick it up on 466 00:28:30,870 --> 00:28:34,787 your desktop. 467 00:28:34,870 --> 00:28:38,927 This cloud harness unlox a lot of possibilities and one 468 00:28:39,010 --> 00:28:41,397 important example that we think you're going to love is our no 469 00:28:41,480 --> 00:28:44,797 curt cloud that we're launching today. 470 00:28:44,880 --> 00:28:48,967 We wanted to give defenders a better set of tools to harden 471 00:28:49,050 --> 00:28:50,897 their flurry and the new cloud 472 00:28:50,980 --> 00:28:56,807 the runs this cloud 473 00:28:56,890 --> 00:29:02,047 environments. We've add some highly requested new features liker about 474 00:29:02,130 --> 00:29:04,517 deduplication, scheduled scans, a new interSanta Fees and you 475 00:29:04,600 --> 00:29:08,057 can start using it today. 476 00:29:08,140 --> 00:29:12,387 Next up we're launching a new AP I canner to make this available 477 00:29:12,470 --> 00:29:16,327 to you. The API includes great 478 00:29:16,410 --> 00:29:20,067 components like the harness, multi-agent controls and much 479 00:29:20,150 --> 00:29:20,267 more. 480 00:29:20,350 --> 00:29:27,350 This is the same technology that 481 00:29:27,990 --> 00:29:28,207 powers dots. 482 00:29:28,290 --> 00:29:33,977 And we're bringing computer use 483 00:29:34,060 --> 00:29:39,447 to API. We want to show you a demo so 484 00:29:39,530 --> 00:29:43,457 say you're building an agent that tests websites this. Agent 485 00:29:43,540 --> 00:29:45,457 can use a computer do open the 486 00:29:45,540 --> 00:29:49,457 browser, click through the 487 00:29:49,540 --> 00:29:53,627 panhandles and test the flow. 488 00:29:53,710 --> 00:29:55,597 We've worked with them on 489 00:29:55,680 --> 00:29:57,737 bedrock managed agents. 490 00:29:57,820 --> 00:30:01,607 You be build open AI agents 491 00:30:01,690 --> 00:30:05,477 acing long side your existing 492 00:30:05,560 --> 00:30:10,077 applications and data. 493 00:30:10,160 --> 00:30:14,257 Okay, here's a big one. When you build these agents nay 494 00:30:14,340 --> 00:30:16,187 need to work safely and protect 495 00:30:16,270 --> 00:30:16,457 your data. 496 00:30:16,540 --> 00:30:23,540 So today we're proving open AI 497 00:30:26,150 --> 00:30:30,067 private tension. Z E-ZD.R. liver delivers 498 00:30:30,150 --> 00:30:36,337 frontier model safety without storing your content on our servers and we can provide 499 00:30:36,420 --> 00:30:39,277 privacy. Together we think these set a 500 00:30:39,360 --> 00:30:43,147 new standards for privacy. 501 00:30:43,230 --> 00:30:46,987 We have worked with our largest 502 00:30:47,070 --> 00:30:49,087 customers to design systems that 503 00:30:49,170 --> 00:30:51,227 meet their highest standards so 504 00:30:51,310 --> 00:30:54,657 they can apply this tension and 505 00:30:54,740 --> 00:30:58,667 you can too. Now, all of this and everything 506 00:30:58,750 --> 00:30:59,297 you're becoming needs both speed 507 00:30:59,380 --> 00:31:00,697 and reliability. 508 00:31:00,780 --> 00:31:06,467 We've been hard alts work on 509 00:31:06,550 --> 00:31:08,637 this. We've still kept more than 510 00:31:08,720 --> 00:31:10,477 99 percent reliability. 511 00:31:10,560 --> 00:31:14,477 We have cut the time to first 512 00:31:14,560 --> 00:31:16,617 token by 45 percent and we will 513 00:31:16,700 --> 00:31:20,487 keep going on this but these 514 00:31:20,570 --> 00:31:22,587 improvements make this the most 515 00:31:22,670 --> 00:31:26,487 reliable API available on the 516 00:31:26,570 --> 00:31:30,767 market today. Thank you. 517 00:31:30,850 --> 00:31:32,767 That's a lot of new things for 518 00:31:32,850 --> 00:31:36,497 developers to build with and it 519 00:31:36,580 --> 00:31:43,580 wouldn't be dev day -- 520 00:31:46,430 --> 00:31:46,607 >> Thanks. 521 00:31:46,690 --> 00:31:52,617 I'll be right there. 522 00:31:52,700 --> 00:31:56,487 Hey codex can you help me? 523 00:31:56,570 --> 00:32:00,497 Can you welcome everyone and 524 00:32:00,580 --> 00:32:07,327 share a fun fact. 525 00:32:07,410 --> 00:32:10,797 >> Hello dev day. Great to have you here and hello to everyone watching around the 526 00:32:10,880 --> 00:32:14,777 world. Fun fact at San Francisco's 1915 527 00:32:14,860 --> 00:32:19,677 world fair visitors could hear voicings from New York over the 528 00:32:19,760 --> 00:32:19,947 telephone. 529 00:32:20,030 --> 00:32:27,030 Today you have an AI on the mic. 530 00:32:33,470 --> 00:32:37,427 There's Romain. 531 00:32:37,510 --> 00:32:41,537 >> Last year we did not have an 532 00:32:41,620 --> 00:32:43,797 alpha codex, now I don't have 533 00:32:43,880 --> 00:32:49,407 enough time today to show you half the things we're shipping 534 00:32:49,490 --> 00:32:49,807 so let's dive in. 535 00:32:49,890 --> 00:32:53,407 It's been awesome to see how 536 00:32:53,490 --> 00:32:59,417 you've been creating 3-D worlds. 537 00:32:59,500 --> 00:33:04,287 What I wanted to try is how question change it in realtime 538 00:33:04,370 --> 00:33:06,257 with UltraFast. 539 00:33:06,340 --> 00:33:13,340 So let's dive in. 540 00:33:34,000 --> 00:33:35,917 We might have to tie it this time. 541 00:33:36,000 --> 00:33:39,927 It's okay. Okay. 542 00:33:40,010 --> 00:33:42,057 You know what I'll just simply 543 00:33:42,140 --> 00:33:43,927 tell you. 544 00:33:44,010 --> 00:33:49,937 Fly me to the stage. 545 00:33:50,020 --> 00:33:53,937 All right. So what's interesting here is 546 00:33:54,020 --> 00:33:54,537 that I'm be the interact with 547 00:33:54,620 --> 00:33:58,077 the room with UltraFast so what 548 00:33:58,160 --> 00:34:02,577 I'll do here is I'll go back here and let's go something else 549 00:34:02,660 --> 00:34:04,217 with this room now. 550 00:34:04,300 --> 00:34:06,317 It feels a little empty so I'll 551 00:34:06,400 --> 00:34:10,057 say why don't you fly the dots 552 00:34:10,140 --> 00:34:15,957 to their seats, for instance. 553 00:34:16,040 --> 00:34:20,127 So we're sending this world in realtime as we go. 554 00:34:20,210 --> 00:34:22,067 Pretty cool. 555 00:34:22,150 --> 00:34:26,037 Maybe we can do something a bit 556 00:34:26,120 --> 00:34:30,077 more risky and why don't we turn 557 00:34:30,160 --> 00:34:32,277 the camera to the room and I'll 558 00:34:32,360 --> 00:34:38,317 say put the live stream on the big screen and let's see what happens this. 559 00:34:38,400 --> 00:34:39,987 There we go. Wow. 560 00:34:40,070 --> 00:34:44,257 So if you want to wave at thed a 561 00:34:44,340 --> 00:34:48,627 conference for those of you watching on uni. Great to have 562 00:34:48,710 --> 00:34:51,997 you. So with UltraFast you can build 563 00:34:52,080 --> 00:34:57,867 and change the world like this or build anything almost as fast 564 00:34:57,950 --> 00:34:58,267 as you can think. 565 00:34:58,350 --> 00:35:02,137 But I want to show you a few -- 566 00:35:02,220 --> 00:35:05,147 so let's switch to the terminal. 567 00:35:05,230 --> 00:35:09,377 Now, we've refreshed the -- with tons of improvementings you've 568 00:35:09,460 --> 00:35:11,347 been asking for. So let's start 569 00:35:11,430 --> 00:35:16,987 with this empty project here on 570 00:35:17,070 --> 00:35:21,097 react. So let's see if I can start voice here. 571 00:35:21,180 --> 00:35:25,297 We might have have some voice difficulties at the moment which 572 00:35:25,380 --> 00:35:28,997 is unfortunate but it's okay. We're going to do ultra fast 573 00:35:29,080 --> 00:35:31,007 anyway but this time we'll do it 574 00:35:31,090 --> 00:35:35,077 with simply the command line, so 575 00:35:35,160 --> 00:35:39,007 we'll say build an app that 576 00:35:39,090 --> 00:35:42,877 picks three attendee at random. 577 00:35:42,960 --> 00:35:47,217 I want do give them free tickets for next year. 578 00:35:47,300 --> 00:35:49,117 So let's see what happens here. 579 00:35:49,200 --> 00:35:55,027 So it would be surfacer and better with voice but -- I wish 580 00:35:55,110 --> 00:35:55,627 I could show it to you but we 581 00:35:55,710 --> 00:36:01,137 have this powered by GPT live. 582 00:36:01,220 --> 00:36:03,297 So this is not dictation, you 583 00:36:03,380 --> 00:36:05,207 can react in realtime. 584 00:36:05,290 --> 00:36:09,037 And boom our app is now ready 585 00:36:09,120 --> 00:36:13,307 with ultra fast. Now, let's do something more 586 00:36:13,390 --> 00:36:15,277 interesting and try switching is 587 00:36:15,360 --> 00:36:19,117 that seven and I'll say make it 588 00:36:19,200 --> 00:36:25,057 six attendees. Why no and boom it's already 589 00:36:25,140 --> 00:36:27,297 changed. So okay, let's bring in full 590 00:36:27,380 --> 00:36:33,067 screen and shall we click on this button? 591 00:36:33,150 --> 00:36:37,267 Let's do it. Legislate see who's coming back 592 00:36:37,350 --> 00:36:39,307 to dev day 2027 now. 593 00:36:39,390 --> 00:36:45,007 Now we have six winners. 594 00:36:45,090 --> 00:36:51,317 Congrats! [Applause.]. My dot will take I want from here and e-mail you these 595 00:36:51,400 --> 00:36:52,887 details later. 596 00:36:52,970 --> 00:36:59,187 So that was UltraFast and with 597 00:36:59,270 --> 00:37:02,727 that -- but let's go back to the an. 598 00:37:02,810 --> 00:37:05,097 Now, I want to show you something cool here but before 599 00:37:05,180 --> 00:37:06,867 we do so I will actually start 600 00:37:06,950 --> 00:37:10,597 this project right now. 601 00:37:10,680 --> 00:37:16,507 We will go back to it in a second, I. to tell you awe Astra 602 00:37:16,590 --> 00:37:18,777 adventures came together. It's a simple idea to build a 603 00:37:18,860 --> 00:37:22,747 game where you can fly into space in 3-D and it started from 604 00:37:22,830 --> 00:37:24,647 this sketch on paper. 605 00:37:24,730 --> 00:37:26,647 Now, few why take a look after a 606 00:37:26,730 --> 00:37:30,517 couple turns we go place but as 607 00:37:30,600 --> 00:37:34,527 you can tell the graphics were a 608 00:37:34,610 --> 00:37:38,527 little rough. So as you can see it's easy to 609 00:37:38,610 --> 00:37:42,637 then after a few turns go to a place where you have this 610 00:37:42,720 --> 00:37:46,967 amazing graphics and go from a sketch to something that is very 611 00:37:47,050 --> 00:37:50,537 beautiful that you can use in a real game. 612 00:37:50,620 --> 00:37:56,547 So we have to refine the game dynamic and it should now be 613 00:37:56,630 --> 00:37:58,817 waiting nor us. Now, instead of playing this 614 00:37:58,900 --> 00:38:00,887 game what I will say is show us 615 00:38:00,970 --> 00:38:04,557 how you play the game. 616 00:38:04,640 --> 00:38:06,657 What's doctoring here is is that 617 00:38:06,740 --> 00:38:09,027 now that we're initiating a use 618 00:38:09,110 --> 00:38:12,567 connection so that's means 619 00:38:12,650 --> 00:38:18,767 thatted Astra will be able to 620 00:38:18,850 --> 00:38:20,937 start the game, and if you take 621 00:38:21,020 --> 00:38:24,977 a look at the screen on the lefe 622 00:38:25,060 --> 00:38:28,417 model is making and on the far 623 00:38:28,500 --> 00:38:34,487 right of the screen you'll see the keys I said expressing on 624 00:38:34,570 --> 00:38:38,657 the game. But I love watching Astra figure 625 00:38:38,740 --> 00:38:40,657 occupant a game that was 626 00:38:40,740 --> 00:38:46,297 designed for us humans. 627 00:38:46,380 --> 00:38:46,727 So you get the idea. 628 00:38:46,810 --> 00:38:50,737 I'll pause it hereful Astra is 629 00:38:50,820 --> 00:38:54,607 also incredible at using any 630 00:38:54,690 --> 00:38:58,447 other app on your commuter. 631 00:38:58,530 --> 00:39:02,647 N. let me show you. If I bring in a different 632 00:39:02,730 --> 00:39:04,617 project here what I can do is 633 00:39:04,700 --> 00:39:08,457 take an app shot and using an 634 00:39:08,540 --> 00:39:12,587 app shot I'll dictate a quick 635 00:39:12,670 --> 00:39:12,887 prompt here. 636 00:39:12,970 --> 00:39:18,697 Audit my app across screen sizes 637 00:39:18,780 --> 00:39:22,467 acing long the way and send this prompt. 638 00:39:22,550 --> 00:39:27,107 Nice. So now that I sent this prompt let's broking back tonight app here. 639 00:39:27,190 --> 00:39:28,707 I'm learning to fly so this is a 640 00:39:28,790 --> 00:39:34,477 little personal project I made. 641 00:39:34,560 --> 00:39:40,487 Now, if you look at what's happening, I love app shots. 642 00:39:40,570 --> 00:39:44,487 They bring the context and in this case it teaches codex to 643 00:39:44,570 --> 00:39:46,487 navigate the computer in this app. 644 00:39:46,570 --> 00:39:48,627 So it starting to click on the 645 00:39:48,710 --> 00:39:52,297 simulator, opening the new app, 646 00:39:52,380 --> 00:39:56,197 figuring out the features and trying everything. 647 00:39:56,280 --> 00:40:00,407 It'ses in the Nateing to watch live but as you're working on 648 00:40:00,490 --> 00:40:04,347 something else you can move on and let it run in the 649 00:40:04,430 --> 00:40:06,207 background. 650 00:40:06,290 --> 00:40:12,217 So now you've seen the app, the CL.I. and I'm excited about how 651 00:40:12,300 --> 00:40:16,217 we've completely reimagined codex in the cloud. 652 00:40:16,300 --> 00:40:16,487 It's fast. 653 00:40:16,570 --> 00:40:20,327 It works with all the same tools 654 00:40:20,410 --> 00:40:24,327 and you can now send your most 655 00:40:24,410 --> 00:40:26,467 ambitious apps on the the cloud 656 00:40:26,550 --> 00:40:32,467 without having to reopen some 657 00:40:32,550 --> 00:40:36,537 for this example length say I want to create a new task but 658 00:40:36,620 --> 00:40:38,577 I'll choose the cloud and I'll 659 00:40:38,660 --> 00:40:40,707 send the prompt, like rewrite 660 00:40:40,790 --> 00:40:46,717 the entire back end in rust. 661 00:40:46,800 --> 00:40:50,687 Why not? Now, this will take a while but 662 00:40:50,770 --> 00:40:54,487 that's great. I know it will get cowboy in the 663 00:40:54,570 --> 00:40:56,797 cloud and I can check oven it 664 00:40:56,880 --> 00:41:00,727 later and do the handoff. 665 00:41:00,810 --> 00:41:06,607 It's been a very popular feature for all of you. 666 00:41:06,690 --> 00:41:06,867 All right. 667 00:41:06,950 --> 00:41:10,607 So zooming of codex we have 668 00:41:10,690 --> 00:41:14,547 incent models. 669 00:41:14,630 --> 00:41:16,917 What's exciting that we build 670 00:41:17,000 --> 00:41:22,517 products like dots and codex 671 00:41:22,600 --> 00:41:26,527 using the same tools. So before I go I wanted to show 672 00:41:26,610 --> 00:41:28,727 you one last demo that brings 673 00:41:28,810 --> 00:41:35,810 all the modalityies together. 674 00:41:38,620 --> 00:41:42,637 Last light I asked codex to wire 675 00:41:42,720 --> 00:41:48,617 everything up here so let's take a look here if we can connect to 676 00:41:48,700 --> 00:41:48,777 it. 677 00:41:48,860 --> 00:41:52,547 All right. 678 00:41:52,630 --> 00:41:56,757 Let's see if it can work over here. 679 00:41:56,840 --> 00:41:58,757 Look at the audience and make a 680 00:41:58,840 --> 00:42:02,627 cool drawing. 681 00:42:02,710 --> 00:42:04,697 All right. 682 00:42:04,780 --> 00:42:08,597 You're in this picture. 683 00:42:08,680 --> 00:42:10,567 Let's see what happens. 684 00:42:10,650 --> 00:42:14,707 It's incredible what you can now 685 00:42:14,790 --> 00:42:18,707 accomplish in codeck. 686 00:42:18,790 --> 00:42:24,577 I add image generation, added 687 00:42:24,660 --> 00:42:28,587 voice intelligence to see entrap this little robot and give it 688 00:42:28,670 --> 00:42:30,927 dozens of capabilities and so 689 00:42:31,010 --> 00:42:32,887 much more we can add. 690 00:42:32,970 --> 00:42:36,897 Hey lavender can you quack if 691 00:42:36,980 --> 00:42:40,597 you can hear me? 692 00:42:40,680 --> 00:42:46,737 Are you excited to be at dev? 693 00:42:46,820 --> 00:42:48,137 [Quacking.] >>. 694 00:42:48,220 --> 00:42:50,107 >> And here's the photo of you 695 00:42:50,190 --> 00:42:57,190 that it just made. 696 00:42:58,070 --> 00:43:02,217 [Applause.]. >>> of course the duck can back 697 00:43:02,300 --> 00:43:05,987 around and do so many other things. 698 00:43:06,070 --> 00:43:08,097 So think the key takeaway here 699 00:43:08,180 --> 00:43:13,997 is this is just a quick demo but 700 00:43:14,080 --> 00:43:16,237 we're excited about using our AP 701 00:43:16,320 --> 00:43:19,307 I canses with hard ware. 702 00:43:19,390 --> 00:43:21,337 They will now be able to do 703 00:43:21,420 --> 00:43:25,307 things like helping robots take 704 00:43:25,390 --> 00:43:29,117 fast actions based on what they see in realtime. 705 00:43:29,200 --> 00:43:32,987 So come see me later if you want 706 00:43:33,070 --> 00:43:35,117 to see that. 707 00:43:35,200 --> 00:43:42,200 We're in an amazing time to be a 708 00:43:43,210 --> 00:43:45,127 builder. 709 00:43:45,210 --> 00:43:48,237 So so much and have a great day, 710 00:43:48,320 --> 00:43:52,167 and back to you, Sam. 711 00:43:52,250 --> 00:43:56,507 >>. >> Thanks, Romain. 712 00:43:56,590 --> 00:43:58,237 That was awesome. 713 00:43:58,320 --> 00:44:04,387 It's amazing what you can build. 714 00:44:04,470 --> 00:44:07,657 We want to help people what you 715 00:44:07,740 --> 00:44:09,617 build and help you build a 716 00:44:09,700 --> 00:44:11,587 business. 717 00:44:11,670 --> 00:44:17,797 Around 1.2 billion people use 718 00:44:17,880 --> 00:44:22,097 ChatGPT. We have try versions of this before but I feel good about our 719 00:44:22,180 --> 00:44:23,397 approach this time because we're 720 00:44:23,480 --> 00:44:25,237 giving you the same tools we 721 00:44:25,320 --> 00:44:32,320 used toe to build ChatGPT. 722 00:44:39,200 --> 00:44:43,317 So weather headline signing with 723 00:44:43,400 --> 00:44:43,957 ChatGPT they can sign into your 724 00:44:44,040 --> 00:44:48,387 app can use their tokens they 725 00:44:48,470 --> 00:44:52,897 use in their ChatGPT plan. This makes I would easier for 726 00:44:52,980 --> 00:44:58,367 people the try your product: they're already paying for an AI 727 00:44:58,450 --> 00:45:00,607 subjection and now you don't have to cover their cost. 728 00:45:00,690 --> 00:45:03,907 We're starting with the 16 729 00:45:03,990 --> 00:45:07,777 launch partners today. 730 00:45:07,860 --> 00:45:09,917 Next we're launching plug in 731 00:45:10,000 --> 00:45:15,787 extensions so the developers can 732 00:45:15,870 --> 00:45:20,127 builded in plug ins. With plug inextensions you be 733 00:45:20,210 --> 00:45:22,197 build an editor, a dashboard and 734 00:45:22,280 --> 00:45:26,927 a work space to work in ChatGPT 735 00:45:27,010 --> 00:45:32,937 and codex. You can see your 736 00:45:33,020 --> 00:45:35,277 upcoming meetings from your 737 00:45:35,360 --> 00:45:36,937 calendar. 738 00:45:37,020 --> 00:45:43,117 You can take notes in ChatGPT. 739 00:45:43,200 --> 00:45:46,947 Here's another example. 740 00:45:47,030 --> 00:45:51,217 You can open a design, see your team's comments and ask ChatGPT 741 00:45:51,300 --> 00:45:53,157 to maybe a revision. 742 00:45:53,240 --> 00:45:56,957 You can work on your designs 743 00:45:57,040 --> 00:45:59,197 alongside your conversation. 744 00:45:59,280 --> 00:46:04,967 Here's another example from 745 00:46:05,050 --> 00:46:08,977 Adobe. We are very excited to see what 746 00:46:09,060 --> 00:46:13,547 you all build. And you can bring these same plugs into ChatGPT sites now 747 00:46:13,630 --> 00:46:14,977 too. 748 00:46:15,060 --> 00:46:18,847 We've seen millions of sites of 749 00:46:18,930 --> 00:46:24,987 Al types get built within a few 750 00:46:25,070 --> 00:46:29,097 months of launch. You can see how can experience 751 00:46:29,180 --> 00:46:31,057 changes based on the user. 752 00:46:31,140 --> 00:46:32,967 It may recollects it easy to 753 00:46:33,050 --> 00:46:38,037 create personal sizeized 754 00:46:38,120 --> 00:46:45,120 software anniversary. 755 00:46:45,990 --> 00:46:51,887 Now we're mixing it so that 756 00:46:51,970 --> 00:46:55,887 plugs can get discovers within the flow of a conversation. 757 00:46:55,970 --> 00:47:00,187 ChatGPT will recognize if your plug in can help and they can connect it right there. 758 00:47:00,270 --> 00:47:04,727 We've also heard your feedback about the plug in submission 759 00:47:04,810 --> 00:47:08,737 process and we're going to simplify that. 760 00:47:08,820 --> 00:47:13,107 You can request a human review without starting the whole submission over. 761 00:47:13,190 --> 00:47:14,867 So finally with all the tot we 762 00:47:14,950 --> 00:47:20,747 want to make it easy I to sell 763 00:47:20,830 --> 00:47:23,177 your product. 764 00:47:23,260 --> 00:47:25,317 We have 30 plus launch partners 765 00:47:25,400 --> 00:47:29,317 today and our enterprise 766 00:47:29,400 --> 00:47:31,557 customers can use their -- this 767 00:47:31,640 --> 00:47:37,597 gives you a are another way to reach customers already working 768 00:47:37,680 --> 00:47:40,667 with us. Through our partnership with 769 00:47:40,750 --> 00:47:42,767 base ten you can access open 770 00:47:42,850 --> 00:47:49,607 source models today. 771 00:47:49,690 --> 00:47:54,077 [Applause.]. >>> thank you. We shipped a lot of stuff this 772 00:47:54,160 --> 00:47:58,077 morning. New ways to work with AI and 773 00:47:58,160 --> 00:47:58,417 dots in space. 774 00:47:58,500 --> 00:48:04,287 Now models with ChatGPT 6.1 and 775 00:48:04,370 --> 00:48:06,227 sol. 776 00:48:06,310 --> 00:48:08,227 Codeck and the cloud. 777 00:48:08,310 --> 00:48:14,097 A lot of new ways to help o you 778 00:48:14,180 --> 00:48:18,007 build a business. It's a big day for us and we're 779 00:48:18,090 --> 00:48:18,637 excited to see what you're going 780 00:48:18,720 --> 00:48:24,007 to do with this. 781 00:48:24,090 --> 00:48:28,117 Every day we hear stories about how you're adding this 782 00:48:28,200 --> 00:48:30,017 technology to do this is that 783 00:48:30,100 --> 00:48:32,417 last year would have in the 784 00:48:32,500 --> 00:48:38,027 meantime out of reach. 785 00:48:38,110 --> 00:48:42,527 And we want to show you a video 786 00:48:42,610 --> 00:48:46,727 of people here with uss dev 787 00:48:46,810 --> 00:48:51,837 day today. So let's meet them. 788 00:48:51,920 --> 00:48:54,137 >> Highlights on. Cameras rolling. 789 00:48:54,220 --> 00:48:58,477 Here's to you. The builders. 790 00:48:58,560 --> 00:49:00,107 The new artists. 791 00:49:00,190 --> 00:49:02,747 The makers of our time. 792 00:49:02,830 --> 00:49:06,417 The momentum generation. 793 00:49:06,500 --> 00:49:12,357 You're not just writing code. You're rewriting what's 794 00:49:12,440 --> 00:49:16,227 possible. Expanding the collective. 795 00:49:16,310 --> 00:49:19,197 Letting intuition guide the way. 796 00:49:19,280 --> 00:49:24,397 You help dream up new businesses 797 00:49:24,480 --> 00:49:26,267 and upcycle old ones. 798 00:49:26,350 --> 00:49:29,877 You're opening the doors of 799 00:49:29,960 --> 00:49:34,007 learning and decoding the immune 800 00:49:34,090 --> 00:49:37,347 system. 801 00:49:37,430 --> 00:49:41,317 What you create, it lives. 802 00:49:41,400 --> 00:49:42,957 It changes. 803 00:49:43,040 --> 00:49:46,627 It moves with us. 804 00:49:46,710 --> 00:49:51,127 Here's to you. 805 00:49:51,210 --> 00:49:51,467 The builders. 806 00:49:51,550 --> 00:49:58,550 The makers of our time. 807 00:50:01,220 --> 00:50:03,307 [Applause.]. >>> we get me to bill AI every 808 00:50:03,390 --> 00:50:06,477 day and that's often but still 809 00:50:06,560 --> 00:50:08,617 one of the coolst things egret 810 00:50:08,700 --> 00:50:12,587 to do is to see what you create. 811 00:50:12,670 --> 00:50:14,757 We have a great rest of the day planned. 812 00:50:14,840 --> 00:50:16,687 We hope you really enjoy it. 813 00:50:16,770 --> 00:50:20,627 Peter will be here for a special 814 00:50:20,710 --> 00:50:22,797 session on how we're continuing 815 00:50:22,880 --> 00:50:26,527 do investor in the open sourcen 816 00:50:26,610 --> 00:50:30,497 and we have a bunch of great 817 00:50:30,580 --> 00:50:32,407 sessions. 818 00:50:32,490 --> 00:50:34,407 And at the end of the day Tejal 819 00:50:34,490 --> 00:50:40,607 and I will be back here. I'm 820 00:50:40,690 --> 00:50:42,547 sure a lot of you have tried something like this but we 821 00:50:42,630 --> 00:50:44,617 wanted to give you a special way 822 00:50:44,700 --> 00:50:48,587 to build with play with what you 823 00:50:48,670 --> 00:50:50,417 build. 824 00:50:50,500 --> 00:50:53,457 So everybody here today is going 825 00:50:53,540 --> 00:50:59,797 to take home a special I 826 00:50:59,880 --> 00:51:03,797 edition -- it's very cool. I think you're going to like it. 827 00:51:03,880 --> 00:51:07,807 You can pick yours up this the mer. 828 00:51:07,890 --> 00:51:09,977 Shop. Finally with all of to new 829 00:51:10,060 --> 00:51:13,807 things we shipped today because 830 00:51:13,890 --> 00:51:20,017 we want you all to big we have 831 00:51:20,100 --> 00:51:26,027 done so many of these that we've 832 00:51:26,110 --> 00:51:30,327 been campaigning to rename the company. 833 00:51:30,410 --> 00:51:31,927 I think he's in the audience 834 00:51:32,010 --> 00:51:35,667 somewhere. 835 00:51:35,750 --> 00:51:42,750 Can we find him. 836 00:51:44,690 --> 00:51:48,117 [Applause.]. >> I think it would beside nice 837 00:51:48,200 --> 00:51:53,447 to do a low key reset here. Maybe we can do it? 838 00:51:53,530 --> 00:51:55,447 Is that a real thing? 839 00:51:55,530 --> 00:52:02,530 >> Shall we just press it. 840 00:52:05,510 --> 00:52:07,767 >> I'm very excited to press this. 841 00:52:07,850 --> 00:52:11,267 I have my laptop ready to go. 842 00:52:11,350 --> 00:52:15,467 Ready the for a count down? 843 00:52:15,550 --> 00:52:17,277 Let's go. 844 00:52:17,360 --> 00:52:24,217 Three, two, one! 845 00:52:24,300 --> 00:52:26,287 >> thank you for that. 846 00:52:26,370 --> 00:52:30,117 That's obviously for everyone in 847 00:52:30,200 --> 00:52:32,427 the world not just people here. 848 00:52:32,510 --> 00:52:34,357 Before we wrap up I want to 849 00:52:34,440 --> 00:52:36,357 share one more thought. 850 00:52:36,440 --> 00:52:40,127 People often talk about AI as an 851 00:52:40,210 --> 00:52:44,437 industrial rev revolution. The 852 00:52:44,520 --> 00:52:50,137 best part of AI is not about 853 00:52:50,220 --> 00:52:52,007 making people cogs in a machine 854 00:52:52,090 --> 00:52:54,147 I believing the future in we get 855 00:52:54,230 --> 00:52:58,147 this right can be more like a 856 00:52:58,230 --> 00:53:02,157 Renaissance than an industrial revolution. 857 00:53:02,240 --> 00:53:04,287 AI should beside about giving 858 00:53:04,370 --> 00:53:06,257 people more power in their lives 859 00:53:06,340 --> 00:53:10,327 and with everything we launch 860 00:53:10,410 --> 00:53:14,167 today we hope we can get do that future. 861 00:53:14,250 --> 00:53:16,767 So thank you all very much for all you do. 862 00:53:16,850 --> 00:53:20,537 Thank you for coming to San 863 00:53:20,620 --> 00:53:22,707 Francisco and for watching 864 00:53:22,790 --> 00:53:29,790 online.