OpenAI's New ChatGPT-4o Destroys Many Startups... Is Yours Dead?

OpenAI just dropped a brand new model, GPT4O.O standing for OmniModal, which means it is encompassing of all types of information, not just text, text, audio and video. Kieran, give everybody watching your just first uninitiated raw thoughts on the news from 24 hours ago.

First of all, for all of you people that are on Twitter and are underwhelmed, you are all wrong. First of all, you're. I suspect you are not going to be impressed until the Terminators knock on your door and ask for Sarah Connor. Right, that's where you are. I'm super impressed. Now I will say, the first five minutes I was like, huh, I don't know. And then what OpenAI do is they suck you in. They do the demos and then they do the biggest Oprah Winfrey, which we will get to of all time, which is you all get this incredible model for free. Or free. The entire Internet should be talking about this. You've got human level intelligence for free. But I thought the, the multimodal people are sleeping on it is a true multimodal model, which means that is multimodal token in, multimodal token in. Which means there's like a singular model that can do text, video and audio. Where has happened in this past is these models pass that around and the results are incredible. And we should start with number one, voice. Like the voice demo we have now, like a Jarvis type experience in our hands, in our pockets.

The what we're showing here, I think is the best description of what AI did. Andre Karpathy, I can never get his name right. I'm sorry, but who's one of the preeminent AI experts in the world? He tweeted, he's like, OpenAI is releasing a combined text audio vision model that processes all three modalities in one single neural network, which can then be real time voice translation as a special use case, as an afterthought. So a lot of the news Yesterday was around OpenAI releasing voice translation. And what he's essentially saying is like, whoa, whoa. It is way better, way more complex, way deeper than that. And I think what you're saying is, yeah, the voice part is magical and we want to start there, but there's a lot of depth actually behind that.

Right. The, the, the power in this is that it's truly multimodal, which we'll get into, which is like voice to text, text to voice, a video to voice, which we saw some demo demos of.

Hold on, I got one other thing before we start, because I do think we should level set for everybody. I sent you this tweet from a guy named Jeff Weinstein that I just thought was really good, right? Which is like, there's no way to watch Those demos from OpenAI and go back to your regularly scheduled day, clap hands and say, okay, team, let's double down on our current linear, hard coded, brittle, single, single language, text only, operational intensive and unscalable approach for solving anything.

Right.

And that I thought was just like the call to action, call to arms for anybody, especially if you're in business, especially if you're in marketing. Because if you go through this week and keep doing things the way you always have, then that's on you. That's your fault, right? Like, you have been given an incredible gift to innovate and do things differently.

We woke up in a new world today and that's why I'm surprised that people are underwhelmed. You now have an AI model that is free. Please recognize the thing is free. The only reason you would pay for chat GPT plus is because you need more capacity. But we'll get into like what the thinking behind that is later, but you have it for free. And then this AI assistant can solve problems through text, visual and audio. It can see things and solve problems for you. We'll get into some of the use cases for marketers. Get into some of the use cases overall. Can do it through audio, can do it to text like on the actual demo. It could tell someone how to do meditation, breathing techniques. By listening to the breath. By listening to the breath and then helping them to level set their breath based upon the information they were giving them. That is a brand new world that we live in. Like you could not do that in the past. Everything was much more robotic. Audio was more robotic, video was more robotic. I think we should just get into like the. To open people's eyes maybe. I'm sure it's been shared all across the web, but we should open. Let's open with like one of the voice demos. All right, let's start off with this.

Little clip, this story, but I want a little bit more emotion in your voice, a little bit more drama.

Got it. Let's add some drama. Once upon a time in a world not too different from ours.

So first of all, Darn Hume, that AI that I have been like telling everyone is so cool. It's like this API that allows you to have empathetic, emotional conversations with a chat across your website. We actually demoed that. We showed that it could the other thing it could do is navigate across your website, which we'll also get into because OpenAI, I think you're going to do that. But like, damn it, OpenAI have like provided you with this capability for free.

I really want maximal emotion, like maximal expressiveness much more than you were doing before.

Understood. Let's amplify the drama. Once upon a time in a world not too, right.

It can basically give you human level voice, but not just that, it actually can dictate. It can actually pick up cues in your voice. Did you see that? It can pick up cues in your voice. It can figure out if you're flustered, it can figure out if you're anxious. It can figure out if you're like in a good mood, bad mood. It can actually have a real conversation with you by level setting against the kind of voice tone that you are using. You can ask it to like communicate with you in any way that you want. And the, the other thing that is actually the most important thing about this is it's instant, right? It's hard to have a conversation when you are waiting on a three to five second delay. And the other thing is you can interrupt it so you can have a real conversation. You can interrupt it, say, hey, no, no, don't give me that, give me this. Which is all the things that we were shown with Hume AI. And the last thing I'll just say is we said this on a previous episode. Some of these AI tools just provide the, give you the ability to some to do some of the best demos you can do. They're just such great demos. The way you can demo these products.

Because they're so interactive, look, they're magical, magical demos. And the emotion and lack of lag time in this new omnimodal model is going to be dramatic for marketing, right? Like one of the things that we've been talking about for a while is like, oh, you could have a omnimodal multimodal website where you can just talk to it and it'll help you find the page that has the thing you're looking for. You could build that. Now that's not like a, oh, we're going to get there. That's like, that is here.

That is here.

That is here. All of your marketing content now can also be localized for a very low cost by default, right? Like you can now operate in multiple different languages. Like that is fricking incredible. All of your customer calls, all of your unstructured data can now be, can now be brought to life in incredible ways and be used to train these voice interactions so that you're going to have much better voice support. Voice, voice sales motions for like a lot. I think especially a lot of the early sales stuff like appointment setting, things like that, that is going to all move to artificial intelligence. And it's pretty amazing the trajectory we're on.

Yeah. I want to show another quick example of that, which was there were so many different versions launched by OpenAI on their Twitter handle. This was another one that you're kind of talking through which is like, you know, you are, you are able to use, you are able to use this technology as a meet in assist. Like you can instantly use it as an ea. Right. So this is the version of the desktop app listening in on a meeting and then you can actually talk back to it and ask it to recap notes and ask it, hey, who actually said that? And who said this?

What's happening is OpenAI is becoming the ultimate platform. Right? Like we thought OpenAI was going. One of the things you and I originally said is that like, oh, OpenAI is better off being the AWS, the ultimate platform for powering the next generation of companies. And they're going to do that. Plus they're going to run just a lot of the core functionality of businesses themselves. They're going to have kind of both lines of business. And I would be remiss, Kieran, before we go into other parts of the news, if we just didn't give Sam Altman the Marketer of the Year award. Now it's May, but I think, I think he largely has it wrapped up. He did two things around this announcement that I thought were pretty magical. One, that guy is the great front runner. He is like, hey, I know Google has Google. I o. I'm going to ship better stuff than they're going to ship at that event and I'm going to ship it a day before and I'm going to get all the buzz and get all the attention. And it had me thinking yesterday, Kieran, I want your opinion before I go to point two. Is like, is the big event announcement model done? Like, is anybody who waits to have a big on the calendar event announcement model, are they just asking for the disruptors to come in and front run them?

Well, I think this is what you're talking through, so I'd love to just quickly show this because it's making your point for you. Well, I guess you can, you can split it either way. So this is from Logan Kilpatrick. He was on the new HubSpot AI podcast. He is next wave was OpenAI. He is now over at Google now. He did a pretty cool thing, right? He had this video ready to go. So what's he doing here? He's basically showing the AI the setup conference and he's asking yeah, those big pretty letters on screen, what do those mean?

Those letters represent Google I.O.

A developer conference held by Google.

So what's happening here? Google are going to launch something very similar today. And what you are saying is well, OpenAI have stole all the buzz because when you've launched the thing and, and then you are launching the company who are launching the same thing a day later are not going to get the same sort of buzz. Right. And Sam Altman has front run Google the entire time. He has to have. I think, I think about this. Sam is an incredibly smart person, an incredible marketer. I think he is laying waste to Google's launch strategy. Yeah, not a great interviewer. People may want to give him a break because you know he is going on the podcast. He's doing an incredible job of saying the same things but he's like much more like a politician. I listened to him on the all in podcast. He has, he's, he's been guarded with his words but yeah, I think he is laying waste to like the traditional way to launch things and it is really hard. Google are going to have to change their. I think they're going to get a little more guerrilla like guerrilla warfare, right? Like they're going to little more straight to the people rather wait than waiting for these large events to get these things in people's hands. Because whoever is first here sucks up, sucks out all, sucks up all the oxygen in the room.

Look, we're going to actually maybe should do a separate show on this Kieran but Google needs to, to take a page out of, out of our friend Mark Zuckerberg's book and go to the people directly. Be more agile. Go with what Zuckerberg did when Apple Vision Pro came out where he did like that raw kind of real video comparison of between Oculus and Apple Vision Pro. Like that worked right? And Google is still stuck in this very corporate comms, corporate communications style of marketing. And what we're trying to tell everybody is for any business the effectiveness of that is going down the hill. One more Sam Altman point because I think it is very important and it's something that literally any marketer can do. The genius of what Sam did was not that he, not just that he front run Google. It's that he set expectations appropriately. And what I mean by that, at the end of last week, OpenAI made an announcement that they were going to have this event and make an announcement. And right off the bat, Sam said, it's not GPT5 and it's not a search engine, because there were lots of speculation that that's what they were going to do. And he told people right off the bat, it's not that. And so he managed expectations because if he hadn't done that and he'd come in and release 4.44. Oh, people wouldn't have been as impressed because they've been like, really? I was waiting for five. I was waiting. I thought you were going to have this search engine killer. What's going on? Literally anyone watching this, you can do that. If you tell your customers what is coming and set the right expectations, then you will meet them where they are and they will be very happy with the thing that you are launching or the thing that you're giving them.

That's a great lesson that Companies like Rabbit R1 and Humane AI Pin completely failed at. That's actually the reason those companies failed in those launches is because they didn't manage expectations. Now those companies are dead in the water. Kip, you might want to leave back your rabbit R1 because that is another category of I'm going to put it.

In a little museum case and sell that thing to an AI museum in like 10, 10 years, baby.

It's like, Submit R1 man death by GPT4. Oh. Because that is a category of startups that OpenAI laid waste to last night because the only logical way that you would ever really use an AI assistant that can do something like those companies we're trying to do, where it can see things, it can be with you at all times, help you figure out problems, provide you with all the things you need in your pocket was to incorporate it into the phone. And that is what OpenAI are clearly doing. But one of the things, one of the quick things I want to show before we move on from voice is I'm so excited by this, which is real time translation. Hi, it is.

I love the result.

Insane.

Doing great.

So I would like you to function as a translator. I have a friend here who only speaks Italian and I only speak English. And every time you hear English, I want you to translate it to Italian. And if you hear Italian, I want you to translate it back to it.

If you were Italian, I would be doing Spanish. Have some humor there, Mike.

She wonders if whales could talk. What Would they tell us?

I mean, going on holiday is going to be ten times better, right?

I'm moving to Spain, baby.

I have been the. The. I love Spain. It's where I love to hang out. I have been pretty lazy about learning Spanish. Now AI has solved my problems. I do not need to go learn Spanish because I can just have it there with me at all times and it can translate. And I suspect what I'll be able to do in a not so distant future is, is have that voice be my voice, right? I'll be able to actually clone my voice for the AI to be my voice. So that is another, I think, really incredible use case. And you would not be able to do that if they had not built a truly multimodal model because the latency would be too high. The reason it works so well now is because it's a multimodal model, which Kip kind of showed you at the start. And there's zero latency or there's no real latency. So you can have a more near instant time conversation with someone through the translation. And I think that is an incredible use case.

Oh, Kieran, by the way, did you see that? That one feature, that translation took $340 million off of a company's market cap?

Duolingo.

Yeah, Duolingo's stock went down by $340 million. Because they're like, well, why are people gonna want to learn a new language as much if they could use a real time translator? Which is fascinating and very true. And there's probably a whole other conversation about what this means for the future of language and how we all communicate. But that's another story for another day. But it's showing that like just one step function in something like this can have a massive impact.

So here's the one that I'm fixated on.

Oh, you're so excited. Walk through.

Really applicable to marketers. So this is basically OpenAI. I'll try to get to a little segment that we can show. But basically this is Khan Academy, which I think we're all fans of, and it's helping students to learn in real time. So the students got the iPad. Now, because you have OpenAI and GPT4O running on the iPad, it can see everything that is being created on there. I want to make sure he understands it himself. And he's here in the room, so you can talk to him directly.

Of course, I'd be happy to help. Let's look at the problem together. Can you first identify which sides of the triangle are the opposite adjacent and hypotenuse relative to angle alpha.

All right, so I'm pretty sure this.

Is the angle alpha right here.

Right? Right. The AI is in real time tutoring the student on the iPad. Now why am I obsessed by this use case? Okay, well, if you think about marketing, a lot of marketing is educational content, right? A lot of marketing is being able to educate your customers on problems that they have and how to solve those problems and provide value through that content prior to them ever becoming a customer of yours. And what you do is you basically extrapolate problems that your product solves and solve them through content before you ever solve them through software. One of my aha moments in AI in terms of educational content, and I think I told the story before was I was trying to figure out how to fine tune llama on my laptop and I wanted to do a couple of specific things with it and I wanted like a very specific set of things in a course. And so I went to what did I do? I did the same thing you would normally do, right? I googled for the course, I YouTubed the course. I looked for the course in all of these kind of maven and all of these other places, but some of them had like elements of what I wanted, but none of them had everything that I wanted. And I was like, wow, can Claude just build me the course? And so I went and told it, hey, I want this kind of course over this many days to answer this questions and I want a certain amount of detail and I want after each lesson I want some homework. And the Claude core course was better than any other course that I could have paid for. And that's actually why I ended up creating the AI course for marketers in a very, in a very similar way because I was like obsessed by how good is AI at actually creating course level content. Now that content is somewhat text based content it can do a really good job of, but it's not like a full interactive course because you can't do voice, you can't do video, it can't see what you're doing. Right? You can't show pictures of the things you're doing and actually be able to find feedback. But now you can. And so my push for marketers and I'm interested to hear what you think is we've moved past like the text based educational content era. Your content really needs to be multimodal, needs to be fully interactive across audio, video and text and really can be done from start to finish by the AI. What the Humans need to do is provide templatized information for the AI to create course in the style that you want. But you should be able to have incredible educational content for your customers. And you really have no cap on that if AI is creating it. And so I think every brand will be able to create some version of an academy that is driven through AI but is specialized through the templates that you give the AI or is like customized through the templates that you give the AI that are unique to your customer.

Look, I think one of the big challenges here is that for the largely the history of humanity it has been about like how you learn. How can you find a teacher who knows the thing and can teach you that thing in a way that works for you? I think that how to learn part of learning is changing dramatically. It is like it is going to be much easier to learn than ever, ever before. So what's going to happen now is the, what you are going to learn about because the point you make Kieran, which I think is the right one, which is like now you can, you can have a teacher on anything, literally anything, anything that's known. So unless you're going off on a discovery mission of some new and unknown thing, any known thing you can have a teacher for and you can have that teacher make a perfect custom course just for you. And that is mind blowing. The rate of human pace and human change is going to be dramatic. And I think if you're a business, this, this manifests in a couple ways. One, you're going to be able to teach your customers everything you want them to know about your product or service in a far better and far more detailed way, in a way that's custom to them. That's going to be job one. The second thing Kieran is one of the biggest challenges of most companies today is training people on their team, especially training sales rep representatives so that everybody kind of operates and responds to the customer in the correct way. You are now just going to have a real life sales tutor like from the jump. If you come in and you join a company and you're a new sales rep for that company, you're just going to have an AI tutor that completely tells you, teaches you how to sell that product very situationally specialized to every prospect. It is going to completely transform how humans interact with each other, especially in kind of sales and commerce situations.

All of these tools, AI products that are building these kind of knowledge, chief knowledge assistants for companies, which is primarily like suck in all of the data that you have and Then be able to either answer a sales reps question, a customer support question, be able to help onboard. These employees, they just got an upgrade overnight. Right now they actually have an incredible tutor that can tutor all of these new employees across voice, text and audio. We've talked about this before, which is the incredible thing about AI is overnight these startups building on these products like there's the pro and the cons. The pro is these companies may lay waste to your startup with these updates. Right? I think last night AI note taken apps laid waste, right? Because I can just do that with GPT 4.0. It can the phone down. There was a tool I was looking at for real life meetings and it was you put the phone down, there's an app, it tra. It hears what's happening and then transcribes that and it provides you the notes and the follow ups. Well, the model. OpenAI have just released a model that can do that for me. Right. It laid waste to AI hardware and there's. And I think we're going to get into it, maybe lay in ways to perplexity AI, which I'm. I'm sad about. But that just the way the cookie crumbles in the, in the world of AI, the, the pro to build on these things is you just get an upgrade, right? You get this huge upgrade. I don't know if you saw the founder of Rewind, he's like hey, our costs have just. The thing we haven't even talked about is the costs are way cheaper. He just says wow, our costs have just gone down by 2x overnight. Here you go customers, here's a rebate, right? Here's cheaper product. And so there's like benefits, there's pros and cons. It's a dicey area. But the. But this is a bit of a tangent but the. But the AI tutor for employees for your customers, like it seems, it feels like to me that is a ubiquitous part of your go to market strategy. Both internal for enablement and external for customer enablement. And I think that there is no reason that your customers should not have an incredible experience in how to use your product.

Yeah, I mean there's so much to this announcement because this Omnimodem model is so shocking from a demo perspective that you kind of. We've spent a lot of the show going into that. One of the things we haven't even covered yet is that GPT4O is twice as fast and twice as cheap as the previous model. So it has gotten 2x cheaper. To do to do 10x more interesting use cases. And so what's interesting here, and this is Sam's like core belief, Sam Altman from OpenAI, is that the access to intelligence is going to get 10x better and 10x cheaper. It's just going to get exponentially better and exponentially cheaper over time. And they're proving that to be true. And he's got a great talk around, like, if you look at human history, this is just the.