Why OpenAI's Announcement Was A Bigger Deal Than People Think

Today on the AI Daily Brief, a deep dive on OpenAI spring update the AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, come check out our discord. You can find the link in the show notes. Quick note before we dive into the episode, I don't want to break up this stream of consciousness with any ads. I do want to shout out that at Super Intelligent. You better believe that we are going to start digging into these new OpenAI updates right about now. I for one am particularly excited to try out these new image generation capabilities that have what appears like it could be incredible ability to include specific text as well as native consistent character generation. And so as always, if you haven't checked out Super Intelligent yet and you want to get your AI learning on, go to BeSuper AI and use code podcast for 50% off your first month. Now let's talk about what the heck OpenAI just announced. Welcome back to the AI Daily Brief. Today is one of those days, kind of the opposite of some of the ones we've had recently where everyone is talking about just one thing. And so instead of doing our whole normal brief then main episode sort of conversation, we are just going to focus on the big thing that everyone is talking about, which is of course OpenAI's spring update. Now this is the event that has been rumored for a couple of weeks. For a while there was speculation that we were going to see a search engine, some sort of competition with Google and Perplexity. But towards the end of last week, as the event apparently got delayed a couple of days, it started to come into view that the most likely candidate was some sort of personal assistant update, particularly around voice features. Now this I believe will go down as one of the most divisive initially product updates that OpenAI has ever released. So what we're going to do on this show is first we're going to talk about what they actually shared and then we'll get into the reactions and why. I think it's actually more significant, not less significant than it seems at first. Right away, the first thing you noticed when it kicked off was that Sam Altman was not the one presenting. I could be totally wrong, but I initially took this as a sign that perhaps it wasn't going to be as big an announcement as we might have thought. Sort of with the idea that they were keeping Sam in the background for the big major updates like GPT 4.5 or GPT 5. Now one of the things that you'll hear A lot throughout this assessment of what happened is that I think that people's expectations or hopes really more than expectations of GPT4.5 or GPT5 colored the way that they received what was actually shared. This is of course, in spite of the fact that OpenAI did make it clear in advance that we were not getting GPT 4.5 or GPT 5 quickly. CTO Mira Muradi honed in on three big pieces of the announcement. First, there was a ChatGPT desktop app. Second, there was an updated ChatGPT UI. And three, and obviously the most important, there was a new flagship model called GPT4O. Basically this was described as GPT4 level intelligence, but faster and with better ways to interact with on OpenAI's website, they call it their new flagship model that can reason across audio, vision and text in real time. The O they write stands for Omni and is a step towards much more natural human computer interaction. It accepts any input as combination of text, audio and image, and generates any combination of text, audio and image outputs. Plus they say it's really fast. It can respond to audio inputs in as little as 232 milliseconds, with an average of 320 milliseconds, which is similar to human response time. In a conversation before they got into the demos, the next part of the announcement had to do with accessibility. Specifically, they said with the efficiencies of GPT4 omni, we could bring this to everyone. What that meant was that free users now have access to a GPT four level model, custom GPTs, the GPT store, basically everything that you were paying for. Before paying users didn't have access to any differentiated technology anymore. Instead they had five times the capacity limits. They also would be first in line for new features, as we saw later in the day as GPT4 started immediately rolling out. And as we'll discuss in a little bit, the improvement in what's available at the free base level is hugely massive. And the only reason I think that it wasn't talked about as such is that the vast majority of people who are spending their time watching an OpenAI product video are probably already springing for the GPT plus account. In other words, the free access part doesn't benefit them, so it's easier for them to overlook the significance in aggregate. We'll come back to that though, in a few minutes. GPT4O was also going to impact the API. Specifically, it was going to make it 50% cheaper, which is obviously a significant change. From there, we got into the live demos of the real time conversational capacity of the ChatGPT app. When Mira Murati asked what's different from the existing voice mode that we have? The presenters answered that you can butt in, whatever without throwing it off, that it has real time responsiveness, that the model picks up on emotion, and that it can generate voice in a wide variety of styles. This emotional awareness is pretty significant. One of the demos that they did was telling a bedtime story, and the two presenters kept asking it to change or modulate its speech based on some new criteria. So first they wanted it to be more dramatic, then even more dramatic, then most dramatic of all, which it did each time very successfully, and then they switched it to dramatic, but in a robot voice, and then they had it sing the end of the story. I will note here that even for people who weren't that impressed with anything else, many had the same thought that cassette AI had when they said, Gotta give GPT4O props. That's the most natural sounding AI voice I've ever heard. Next up, they showed off the new vision capabilities. First they did a linear equation where they asked ChatGPT to help walk them through how to solve it. So instead of just pointing the screen at an equation on a piece of paper and asking it to solve it, the presenters were really using it as a tutor more than anything else. And in that way, I think it reflected what they were really showing off, which is these features as not somehow standalone, but as part of a complete assistant experience. And speaking of that assistant capability, they also did a demo where they brought up the ChatGPT desktop app, specifically the conversational version of it, the and were able to ask it about the code that they were writing in a different application simply by copying it into the ChatGPT window. They also showed off ChatGPT describing what it saw on screen after the code was run. The two other demos they did, theoretically from audience input, were real time translation, where one of the presenters spoke in English and then Miro responded in Italian with ChatGPT operating as the translator in real time. And then finally they asked ChatGPT to recognize the emotions looking at someone's face. And then that was it. It was a tight half an hour. There was no big one more thing, Steve Jobs type of moment. And like I said, there were a lot of kind of underwhelmed responses. Abacus AI CEO Bindu Reddy writes, is this me or was that it? What even that was the single most underwhelming thing I've seen this year. I'm not sure what's cool about this. That Google duplex demo from 2019 was way better. The only highlight, if any, was the tone modulation, which wasn't even that spectacular. Theo Jaffe writes, maybe I'll be crucified for this, but I actually wasn't blown away by this demo like I was for the releases of ChatGPT and GPT4. This seems more like a product update than a foundational new capability breakthrough. On the flip side, you add folks like Pete from the neuron who wrote GPT4O is magical. Absolutely magical. Rory wrote blown away that more people aren't blown away. We just went from smartphone to iPhone. Chris France writes LOL New OpenAI model is better than all existing models at everything, supports real time vision and audio and is free what Here's a story that is all too common. You your business gets to a certain size and the cracks start to emerge. Things you used to do in a day are taking a week. You have too many manual processes. You don't have one source of truth. If this is you, you should know these three 37,025 one 37,000. That's the number of businesses which have upgraded to NetSuite by Oracle. NetSuite is the number one cloud financial system streamlining accounting, financial management, inventory, HR and more. 25. NetSuite turns 25 this year. That's 25 years of helping businesses do more with less, close their books in a day, not weeks and drive down costs. One because your business is one of a kind. So you get a customized solution for all of your KPIs in one efficient system with one source of truth. Manage risk, get reliable forecasts and improve margins. Everything you need to grow all in one place. Listen, I know a lot of you guys are entrepreneurs out there and one of the things that is a reality of entrepreneurship is that almost from the moment that you've started, you have more information in more places than you possibly know what to do with. That gets nothing but harder the bigger that you grow. And frankly, right now with AI, companies are growing bigger faster than ever before. If you're facing any of these challenges right now, you can download NetSuite's popular KPI checklist, designed to give you consistently excellent performance absolutely free@netuite.com breakdown that's netsuite.com breakdown to get your own KPI checklist. Netsuite.com breakdown as a listener of this show, I have a strong feeling you like to stay up to date on all things artificial intelligence, including its impact on the workforce. Which is why I highly recommend checking out Managing the Future of Work, the chart topping business podcast from Harvard Business School. HBS professors Bill Kerr and Joe Fuller talk to business leaders, technologists and policymakers grappling with the forces like AI, globalization and demographic shifts that are reshaping the nature of work. Recent guests include IBM Chro Nicola Moreau on how Big Blue is adopting AI, Morningstar CEO Kunal Kapoor on how AI can raise the investment iq, Microsoft corporate Vice President Jared Spitero on how the tech giant is experimenting its way from AI assistance to autonomous agents and many other prominent movers in business and the workforce ecosystem. So don't miss out. Follow Managing the Future of Work on Apple Podcasts, Spotify or wherever you're listening now. But what about the team at OpenAI? What story were they trying to tell? Well, Sam Altman wrote it up explicitly on his blog. He said that he wanted to highlight two parts of the announcement. First, he said, a key part of our mission is to put very capable AI tools in the hands of people for free or at a great price. I'm very proud that we've made the best model in the world available for free in ChatGPT without ads or anything like that. Our initial conception, he continues, when we started OpenAI, was that we'd create AI and use it to create all sorts of benefits for the world. Instead, it now looks like we'll create AI and then other people will use it to create all sorts of amazing things that we all benefit from. We are a business and we'll find plenty of things to charge for and that will help us provide free, outstanding AI service to hopefully billions of people. Second, Sam writes, the new Voice in Video Mode is the best computer interface I've ever used. It feels like AI from the movies, and it's still a bit surprising to me that it's real. Getting to human level, response times and expressiveness turns out to be a big change. The original ChatGPT showed a hint of what was possible with language interfaces. This new thing feels viscerally different. It is fast, smart, fun, natural and helpful. Talking to a computer has never felt really natural for me. Now it does. As we add optional personalization, access to your information, the ability to take actions on your behalf, and more. I can really see an exciting future where we are able to use computers to do much more than ever before. And so I think Sam is getting here at two of the three biggest parts of the the transformation that this represents when you make it free and OpenAI's bet on a new mode of human computer interaction. I'm going to talk about each of those in some more detail, but the third that I want to point out is truly native multimodality of this this was an announcement that was not for a technical audience, at least it didn't seem to be to me. All of it was incredibly simple language and they didn't even show off some of the capabilities. In fact, because they didn't explain it, some people questioned what was going on underneath the hood. Andrew Gao writes for my technical audience. Thoughts on what's behind GPT4O? Is it really multimodal and not converting things to text? That is, you can replicate the demo by using Whisper to convert speech to text, use regular GPT4 and then convert the response to speech using 11 labs. It would be entirely different if OpenAI was actually going from audio waves to audio waves end to end without other models in between. Definitely possible, and would explain the ability to understand and hear breathing in the demo. But this is also doable without that necessarily. Well, Andrej Karpathy, previously of the founding team of OpenAI explained it this way. He said they are releasing a combined text audio vision model that processes all three modalities in one single neural network which can then do real time voice translation as a special case afterthought if you ask it to. In other words, yes, this is true native multimodality. It is not taking language tokens and then converting them. Will De Pooh, who works on video generation at OpenAI, says I think people are misunderstanding GPT4O. It isn't a text model with a voice or image attachment. It's a natively multimodal token in multimodal token out model. You want it to talk fast, just prompt it to need to translate into whale noises. Just use few shot examples. An example that he showed was character consistent image generation just by conditioning it on previous images. He then showed an example and if any of you have spent any time trying to get consistent characters with workarounds like style reference on midjourney or creating a custom GPT as I've done, or using a third party application like Scenario gg, the fact that it might just natively have these capabilities is pretty significant. So to me the three biggest parts of this announcement were one the fact that this best in class model was free for everyone, 2 the fact that it was truly natively multimodal and 3 the fact that OpenAI was clearly making such a huge bet on this new type of human computer interaction as the future of how we interact with AI, but what about when people started to get their hands on it? How are the reactions then? Well, Sully Omar From Cognosis writes, GPT4O is way way faster than GPT4. It feels like an entirely different model. Insanely fast. Andrew Gao again writes to everyone disappointed by OpenAI today, don't be. The livestream was for general consumer audience. The cool stuff is hidden on their site. Some of the examples he gives are text to 3D, hugely advanced text and AI generated images. Andrew points out they're so confident in their text image abilities that they can create fonts with GPT4O and a bunch of other huge things as well. Sully again writes, okay, I get where ChatGPT is going. Ultimate workflow equals screen share with ChatGPT. ChatGPT operates the computer for you. You can interject chat all through voice. It's like having someone there directly working with you. In fact, right now as we're recording this streaming live on X is someone coding in cursor with GPT4O basically as a live coding companion. Others pointed out that the timing of this was no accident. Robert Scoble writes, what was just announced by OpenAI was designed to blunt attacks by Apple and Google as both companies are about to change their voice assistants to LLM based systems that will fix most of the things we hate about both. Apple has lots of advantages that it can brag about like you'll be able to change the brightness on your phone by talking to Siri or be integrated into Apple's ecosystem, that is can you put something on my reminders Apple? Others pointed out that the ChatGPT demo today was basically the demo that everyone freaked out about from Gemini Ultra back in December that then everyone found out was edited to death and not actually representative of its true capabilities. Even more than that though, Google I O is happening tomorrow and Logan Kilpatrick, who notably used to work at OpenAI, shared a video of what is presumably a Gemini assistant looking at the IO stage and explaining it to the person holding the phone. So it seems highly likely that tomorrow we're going to be having a very similar conversation comparing to whatever they announce at Google I O to what we got from OpenAI today. Oh, and as one fun little aside, they did confirm that the I'm also a good GPT2 chatbot that everyone has been freaking out about on Limsys is indeed a version of GPT4O that they've been testing when it comes to real world response. Certainly the real time translation demo seems to have had an impact. One little coder pointed out a 5% drop in Duolingo's price in the wake of the demo. Siqi Chen summed up where I think a lot of people will end up in the long run when he wrote, this will prove to be, in retrospect, by far the most underrated OpenAI of event ever. He even went further and said, TL Dr. GPT4O is a significantly larger improvement over GPT4 than 3.5 was over 3 GPT4O equals GPT 4.75. I think the point here, one that will ultimately be proven out or not by our interactions with it, is that this native multimodality plus the ability to input on the basis of vision and video transforms the use cases of ChatGPT in a huge way that we're probably underestimating initially. Another part of this though, was summed up by Aaron Levy from Box. He wrote the productivity unlock for humanity is pretty insane when AI can bring this level of intelligence to anyone. Like I said, I think the reason that we're not talking more about just how significant the free shift is is that most of us who are doing the talking right now have been paying for ChatGPT since the moment we could. Giving billions of people access to that, though for free is just likely to have an enormous, enormous impact on work, society and everything in between. Ultimately, we'll see. I think it is in no way guaranteed that the way that people will want to interact with these technologies is through these sort of chat modalities or interactions with video. The real world will show us that one way or another. Regardless of what plays out though, it's pretty clear that OpenAI believes that this is truly the future of interaction with AI. I think just because Sam Altman wasn't doing the presentation, just because they might have rushed this a little to get in ahead of Google I O, and just because they didn't announce formally 4.5 or GPT5. It would be a mistake to underestimate how significant this update is in the minds of OpenAI themselves. However, there is going to be a lot more to discuss with this, especially with Google I O coming tomorrow. So that is going to do it for this edition of the AI Daily Brief. Until next time, peace.