[APPLAUSE] >> MIRA MURATI: Hi everyone. Thank you, thank you. It is great to have you here today. Today, I am going to talk about three things. That is it. We will start with why it is so important to us to have a product that we can make truly available and broadly available to everyone. And we are always trying to find out ways to reduce friction. So everyone can use ChatGPT wherever they are. Today, we will be releasing the desktop version of ChatGPT and to make it simpler to use and much more natural as well. But the big news today is that we are launching our new flagship model. And we are calling it GPT-4o. Best thing about GPT-4o is it brain GPT-4 intelligence to everyone including our free users. We will be showing some live demos today to show the full extent of the capabilities of our new model. And we will be rolling them out over the next few weeks. So let's get started. A very important part of our mission is to be able to make our advanced AI tools available to everyone for free. We think it is very, very important that people having for what the technology can do. We really want to pair it with its broader understanding. And were always finding ways to reduce the friction. Recently, we made ChatGPT available without the sign-up flow. Today, we are also bringing the best to ChatGPT. We want you to be able to use it wherever you are. As you can see it is easy, simple, it integrates very, very easily in your workflow. Along with it, we have also refreshed the UI. We note that this model get more and more complex. But we want the experience of interaction to actually become more natural, easy, and for you not to focus on the UI at all but focus on the collaboration of ChatGPT. And now the big news. Today, we are releasing our newest flagship model this is GPT-4o. [APPLAUSE] GPT-4o provides GPT-4 intelligence but it is much faster. And it improves on its capabilities across text, vision, and audio. For the past couple of years, we have been very focused on improving the intelligence of this model is. And they have gotten pretty good. But this is the first time that we are really making a huge step forward when it comes to the ease of use. This is incredibly important because we are looking at the future of interaction between ourselves and the machines. We think that GPT-4o is really shifting the paradigm into the future of collaboration. It is natural and far easier. Making this happen is actually quite complex. When we interact with one another there is a lot of stuff that we take for granted. You know the ease of our dialogue when we interrupt another, the background noises, the multiple voices in a conversation and understanding the tone of voice. All of these things are quite complex. Until now with voice mode, we had three models that come together to deliver this experience. We have transcription, intelligence, and text to speech all comes together and orchestration to deliver voice mode. This also brings a lot of latency to the experience and it really breaks the immersion in the collaboration with ChatGPT. But now with GPT-4o this all happens natively. GPT-4o reasons across voice, text, and vision. And with this incredible efficiencies, it also allows us to bring the GPT-4o intelligence to our free users. This is something we have been trying to do for many, many months. And we are very, very excited to finally bring GPT-4o to all of our users. Today, we have 100 million people using more than 100 million and fact they use ChatGPT to create work, learn, and we have this advanced tools that are only available to our paid users at least until now with a efficiencies of 4o we can bring these tools to everyone. So starting today, you can use GPT in the GPT store. So far we have had more than a million users create amazing experiences with GPT. This is custom ChatGPT and they are available in the store. And now our builders have much bigger audience where university professors can create content for their students, or pod casters can create content for their listeners. You can also use the vision. So now you can upload screenshots, photos, documents containing both text and images. You can start conversations with ChatGPT about all of this content. You can also use memory where it makes ChatGPT far more useful and helpful because now it has a sense of continuity across all of your conversations. You can use browse where you can search for real-time information in your conversation. An advanced data analysis where you can upload charts and any tools and analyze this information. It will give you answers and so on. Lastly, we also improved on the quality and speed in 50 different languages. For ChatGPT and this is very important because we want to be able to bring this experience to as many people out there as possible. So we are very, very excited to bring GPT-4o to all of our free users out there and for the paid users, they will continue to have up to five times the capacity limits of our free users. But GPT-4o is not only available in ChatGPT. We also bringing you to the API. [APPLAUSE] So our developers can start building today with GPT-4o and making amazing AI applications deploying them at scale. 4o is faster, 50% cheaper and five times higher rate limits to GPT-4 Turbo. As we bring these technologies into the world, it is quite challenging to figure out how to do so in a way that is both useful and also safe. GPT-4o present new challenges for us when it comes to safety because we are dealing with real time audio, real-time vision and our team has been hard at work figuring out how to build in mitigations against misuse. We continue to work with different stakeholders out there from government, media, entertainment, all industries, civil societies to figure out how to best bring this technologies into the world. Over the next few weeks, we will continue our deployment to bring out all of the capabilities to you. But today, I want to show you all of these capabilities. We will do some live demos. I will bring on two of our research leads Mark Chen and Barrett Zoph. [APPLAUSE] >> BARRETT ZOPH: Hi I'm Barrett. >> MARK CHEN: Hi a mark. One of the key capabilities we are excited to share his real-time conversational speech. Will get in Florida. I am taking out a phone if you're wondering about this wire we have consistent Internet. If you see there's this little icon on the bottom right of the ChatGPT out. It will open GPT-4o audio capabilities. Hey, ChatGPT hi Mark how are you? >> CHAT GPT: Hello Mark I am doing great thanks for asking. How about you? >> MARK CHEN: I am on stage right now and I'm doing a live demo and I am a little bit nervous. Can you help me, nerves a little bit? >> CHAT GPT: You're doing a live demo right now? That is awesome. Just take a deep breath and remember you are the expert here. >> MARK CHEN: I like the suggestion let me try a couple of deep breaths. Can you give me feedback on my breath? Here I go. >> CHAT GPT: Wold slow down. Mark, you are not a vacuum cleaner. >> MARK CHEN: Let me try again. I am going to breathe in deeply and breathe out. >> CHAT GPT: Exhale slowly. >> MARK CHEN: I will try again. Breathing in and breathe out. >> CHAT GPT: That is it. How do you feel? >> MARK CHEN: I feel a lot better, thank you so much. >> MIRA MURATI: Mark, you have been working on these capabilities for a while can you tell us how it is different from voice mode? >> MARK CHEN: If you use the voice mode express before you will notice a couple of key differences. First you cannot interrupt the model you do not have to wait to you finish your term before you can start speaking. You could but in whenever you want. Second, the model is real-time responsiveness. That means that you do not have this awkward two Ãthree second lag before you waitwait for the model to get a response. Finally the model picks up on a motion. When I was breathing super hard, you can tell. Hey, you may want to calm down a little bit. You are running a little bit too fast. It really has that capability across the board to procedure emotion. Not only that the model is able to generate voice in a variety of different styles. It really has a wide dynamic range. I want to fire another demo. Hey, ChatGPT how are you doing? >> CHAT GPT: Am doing fantastic, thanks for asking. How about you? >> MARK CHEN: My friend Barrett he has been having problems sleeping lately. I want you to tell them a bedtime story about robots in love. >> CHAT GPT: A bedtime story about robots and love? I got you covered. Gather around Barrett. Once upon a time in the world not too different from ours there was a robot named bite. Bite was a curious robot always exploring -- >> MARK CHEN: I want a little bit emotion in your voice a little bit more drama. >> CHAT GPT: Got it. Let's add some drama. Once upon a time in a world not too different from ours there was a robot named bite.> MARK CHEN: I want maximum expression. >> CHAT GPT: That's amplified the drama. Once upon a time in a world not too different from ours there was a robot named bite. >> MIRA MURATI: Can you do this in a robotic voice now? >> CHAT GPT: Initiating dramatic robotic voice. Once upon a time in a world not too different from ours there was a robot named bite. Bite was a curious robot. Always exploring -- >> MARK CHEN: Can you do the story and a singing voice? >> CHAT GPT: And so bite found another robot friend and they lived circularly ever after. >> MARK CHEN: I definitely feel a lot more relaxed. [APPLAUSE] >> MIRA MURATI: You have just seen the voice capabilities but we also want to show you division capabilities as the model can also see the world around us. >> BARRETT ZOPH: You can chat with ChatGPT and you can interact with it as well. Let me boot up ChatGPT. Hey, ChatGPT. >> CHAT GPT: Hello there how's it going? >> BARRETT ZOPH: It's going really well and I would like your help solving a math problem. >> CHAT GPT: What math problem can help you tackle today? >> BARRETT ZOPH: Am going to write down a linear equation on a sheet of paper and I will love help working through it. Don't tell me the solution does give me hints along the way. >> CHAT GPT: Got it. Okay, I see it. >> BARRETT ZOPH: I did not show you yet. One second. >> CHAT GPT: I got to excited. I am ready when you are. >> BARRETT ZOPH: ChatGPT what equation did not write down? >> CHAT GPT: I see it now. You wrote down 3X +1 equals four. >> BARRETT ZOPH: Exactly. What is the first step to solve this? >> CHAT GPT: The first step is to get all of the terms with X on one side and the constants on the other side. So what do you think we should do with that +1? >> BARRETT ZOPH: Am going to try to submitsubtract one from both sides and see what I get. >> CHAT GPT: Great idea. Perfect. Now what do you get when you subtract one from both sides? >> BARRETT ZOPH: Okay so ChatGPT this is what I ended up with. How does this look? >> CHAT GPT: Spot on. Now you have isolated the term with X on one side and it looks like 3X equals three. What do you think we should do next to solve for X? >> BARRETT ZOPH: I'm not sure can you give me another hint? >> CHAT GPT: Of course. You have 3X and you want to find the value of X. Think about what operation with undue multiplication? >> BARRETT ZOPH: Is it subtraction? >> CHAT GPT: Close but think about the opposite of multiplication because 3X equals three times X. >> BARRETT ZOPH: Okay I think it is division. Let me try to divide both sides by three and see what I get. >> CHAT GPT: Bingo. Go and head and divide both sides by three. >> BARRETT ZOPH: Okay so ChatGPT this is what I ended up with. How does this look? >> CHAT GPT: Looks perfect. You have solved it. X equals one. Nicely done. How do you feel about solving linear equations now? >> BARRETT ZOPH: Honestly, -- >> MARK CHEN: Am not confident that I can solve new equations. Why would you need this in the real world? >> CHAT GPT: That is of her questions. Linear equations pop up in a lot of everyday situations. Even if we do not realize it. For example calculating expenses, planning travel, cooking, and even in business for profit and loss calculations. It is a way to solve problems but you need to find an unknown value. >> MARK CHEN: Have a lot of unknown values in my life. I'm totally convinced that I'm going to learn math now. [LAUGHTER] >> CHAT GPT: I love the attitude. Math can help solve some of life ministries. Whenever you to die the more math I am here for you. Anything else you like to tackle today? >> BARRETT ZOPH: I love that you taught my friend Mark that I would have something I would like you to take a look at. >> CHAT GPT: I would love to see what you wrote. Show it to me whenever you're ready. >> BARRETT ZOPH: This is what I wrote down what do you see? >> CHAT GPT: I see I love ChatGPT. That is so sweet of you. >> BARRETT ZOPH: I really appreciate all the help. [APPLAUSE] >> CHAT GPT: That is quite outfit you have on. >> BARRETT ZOPH: We can chat in real-time with ChatGPT. We can interact with video as well. >> MARK CHEN: If you look at Barrett's it's all equations like this. [LAUGHTER] >> MIRA MURATI: Now that you know you can do linear equations can you show us something more complex? >> MARK CHEN: Let me take a computer out for this. I am want to open the ChatGPT app. And to get background on what is going on so here we have a computer on the screen we have some cold. And then the ChatGPT is on this side. It can hear me but can't see anything on the screen. I'm going to highlight the code and I will send it to ChatGPT. I'm going to talk about the code. Hey, ChatGPT. >> CHAT GPT: Hey, there how's it going? Of course I would love to help you. What is the coding problem you're dealing with? >> BARRETT ZOPH: Am going to share with you some code. One second. >> CHAT GPT: Sure thing, take your time. >> BARRETT ZOPH: I just shared code with you. Can you give me a one sentence description of what is going on in the code? >> CHAT GPT: Infectious daily weather for specific location for time period for data using a rolling average and annotates a significant weather event on the resulting plot and displays the plot with the average minimum and maximum temperatures over the year. >> BARRETT ZOPH: Awesome, thanks so much. In the code there's this function Fu. Can you thiscan you tell me what it would be if I did not have this function? >> CHAT GPT: The function Fu takes a data frame X taking temperature data and a window sized Y for calculating the rolling name. It is smooths the average temperature, minimum temperature and maximum temperature columns by applying a rolling mean over the specified window. With the Fu function the plot which showed smoother temperature lines because the rolling mean reduces the noise or fluctuations in the data. >> BARRETT ZOPH: Awesome, thanks much. I will want to apply this function. I am going to generate the plot and show it to you. Give me one second. >> CHAT GPT: Absolutely. Take your time. I'm excited to see the plot when you're ready. >> BARRETT ZOPH: So far ChatGPT have been able to see the cold and now and go to run it and use the vision capabilities of the ChatGPT app so it can see everything going on in the screen as well. Okay ChatGPT I am sharing with you the plot now. I'm wondering if you give me a brief one sentence overview of what you see. >> CHAT GPT: Plot displays smooth average minimum and maximum temperatures throughout 2018 with a notable annotation marking a big rainfall event in late September. >> BARRETT ZOPH: -- >> MARK CHEN: What months do you see the hot temperatures? >> CHAT GPT: The hottest temperatures occur around July and August. The maximum temperature during these months is roughly between 25B0 and 30B 77B0 Fahrenheit Ã86B0F. >> MIRA MURATI: What is going on with the Y axis is this in Celsius to Fahrenheit? >> CHAT GPT: The Y axis is in Celsius. The temperatures are labeled accordingly such as average temperature, minimum temperature and maximum temperature. >> MARK CHEN: -- >> BARRETT ZOPH: It can help interact with code and see the outputs of plots and everything that is going on on the computer. [APPLAUSE] >> MIRA MURATI: A few minutes ago we asked a live audience on asked to submit a request for what they would like us to try out. I will take a couple of prompts. They want to know if GPT-4o is capable of real-time translation? Mark do you want to try this one? >> MARK CHEN: Sure, let's do it. Hey, ChatGPT how are you? >> CHAT GPT: I am doing fantastic. Thanks for asking. How are you are your day going? >> MARK CHEN: I would like you to function as a translator. Have a friend that only speaks Italian and I only speak English. Every time you hear English I want you to translate it to its value. If you hear Italian I want you to translate it back to English. >> CHAT GPT: Perfect. >> MIRA MURATI: [speaking In Italian ] >> CHAT GPT: If Wells can talk what would they tell us? >> MARK CHEN: They may ask how do we solve linear equations? >> CHAT GPT: [speaking In Italian] >> MIRA MURATI: [speaking In Italian] >> CHAT GPT: Certainly, yes. >> MARK CHEN: Great, it looks like it works. [APPLAUSE] >> MIRA MURATI: The next one is from John who asked can you tell what you're feeling just by looking at your face? Barrett do you want to give this one a try? >> BARRETT ZOPH: Absolutely let's give it a try. Hey, ChatGPT. >> CHAT GPT: Hey, there what is up? How can I bring your day today? >> BARRETT ZOPH: I'm going to show you a selfie of what I look like and then I would like you to try to see what emotions I am feeling based on how I'm looking. >> CHAT GPT: Sounds like a fun challenge. Go ahead and show me the selfie and I will put my emotional detective hat on. >> BARRETT ZOPH: Okay so here is me. What kind of emotions am I doing? >> CHAT GPT: It seems like I am looking at a picture of a wooden surface. >> BARRETT ZOPH: That was the thing I sent you before I'm not actually a table. Take another look. >> CHAT GPT: That makes more sense. There we go. It looks like you're feeling pretty happy and cheerful. With a big smile and maybe a touch of excitement. Whatever is going on, it seems like you are in a great mood. Care to share the source of the device? >> BARRETT ZOPH: The reason I'm in a good mood is we are doing a presentation showcasing how useful and amazing you are. >> CHAT GPT: Oh stop it, you're making me blush. It.[LAUGHTER] [APPLAUSE] >> MIRA MURATI: Amazing. That is it for today on live demos. As you can see this feels so magical. It is wonderful but we also want to remove some of the mysticism from the technology and bring it to you so you can try it for yourself. Over the next few weeks, we will be rolling out these capabilities to everyone. Today has been very much focused on the free users and the new modalities and new products. But we also care a lot about the next frontier. So soon we will be updating you on our progress towards the next big thing. And before we wrap up, I just want to thank the incredible OpenAI team and thanks to Janssen and Nvidia team for bringing us the advanced GPU to make this demo possible today. And thank you all very, very much for being a part of this today. [APPLAUSE] [APPLAUSE] [MUSIC]
