OpenAI’s Big Week, Did Google Kill The Web, AWS CEO Resigns

0:00OpenAI had a massive week with the release of a new model, a partnership with Apple, and the departure of its chief scientist. Did Google just kill the web? And the AWS CEO has resigned. All that and more coming up right after this.

0:36your podcasts.

0:43Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional cool-headed and nuanced format. We have a massive week of AI news and news across the tech world to talk about, including the resignation, big resignation inside Amazon, but also these massive new releases from OpenAI and Google. And we are just off of Big Technology's first live event with BAC CEO Aaron Levy. So thanks to everybody who came out last night. The recording is going to be out on the feed next week and it sounds great. And it was just such an energizing event and awesome to have so many of you there.

1:18Awesome to have you there, Ranjan. And speaking of which, let's introduce Ranjan Roy. He writes margins on Substack. Ranjan, welcome to the show. It was good to see you in person, Alex, and not just on the other side of a screen. And it was a fantastic event. And I'm glad this is going to be coming out before. So some of the things Aaron said that were very smart, I can now repeat them and sound smart myself. Yeah, we'll tease them and then we'll talk a little bit more about what's coming because I thought that this show will talk more about the nuts and bolts, the announcements and sort of the implications for the companies.

1:50And then next week on Wednesday, Aaron and I will talk more broadly about the AI space in general. And we also have some questions and some fun segment where we read his tweets. It's really a great one. Let's begin talking about the big news that started the week. I won't tell too much on this, but OpenAI, of course, released GPT-4-O, and O stands for Omni, and this is according to TechCrunch. That refers to the model's ability to handle text, speech, and video. It's going to roll out iteratively across the company's developer and consumer-facing products over the next few weeks. It can do things like you ask it a question and interrupt it.

2:26It will have real-time responsiveness. and it also this is kind of the thing that to me is most intriguing is it upgrades the vision qualities so you can ask it what's going on in the software code by pointing a camera at something you can say what brand of shirt is this person wearing you can even and this is unbelievable you can have it watch a live sports game with you and then explain the rules to you as you go and i think this stuff where you like point the model point your camera with the chat GPT baked in and get a chance to speak with it about what you're seeing really is going to be some fascinating technology that we're going to be able to work in and think through and help take this AI field to the next level.

3:07The question is, when is it going to come out and how much does the reality mirror what we saw in the demos? Because I and many other people tried to use this the day after the event, not available. It's still not available to me later in the week. and we're going to see whether the promise actually meets or the reality actually meets the promise that OpenAI delivered. So Rajan, I'm curious to start off, what was your reaction to these announcements and do you think that there was an overpromise here? Where do you think this is going to leave us? All right, I am very excited by what OpenAI announced and I do not think they're overpromising because I've actually been using the voice interface with the ChatGPT iOS app for a while.

3:50In fact, one of my favorite things I discovered, and this is six to eight months ago on the pre-GPT 4.0 versions, is on a long car ride, I'll ask, you know, with my five-year-old son in the back, tell me a story about and add his name and make it super personalized and give details about our house or his school. And it will give this like 10, 12 minute story. And he's completely amazed and immersed at the fact that how personalized it is. So already the voice interface has been pretty advanced. And now actually having an LLM built specifically around multimodality and incorporating voice, being built around voice, video, image recognition, I think is going to be able to deliver it.

4:40And the thing I'm most excited about is I actually didn't want GPT-5. I thought getting another massive increase in compute power that could potentially solve unimaginable problems of such scale is not what I wanted. I want to see changes in UI. I actually think the biggest limitation with all products right now across Google, Perplexity, OpenAI, is it's just a chat-based interface. And that's limiting. And that only addresses specific use cases and moments. So this really tackling the user interface side of things, to me, was more exciting than if they had launched some big new model that they said gets us closer to sentience and AGI or whatever it would be.

5:28I think this is good. But here's my big question is how are people going to interact with this thing? I mean, we've already seen that voice assistants haven't taken off in the way that we anticipated. And of course, this is due to limitation in technology, but also just in terms of the user interface, right? People don't really love speaking to these things. And then there's the case of the video, right? And yes, the video can look at you and tell you how you look, but it seems much more useful pointed out than pointed in. And what I mean by that is it seems like we might really need smart glasses for this stuff to actually sing as opposed to using it on our phone.

6:03So you're obviously excited about the new formats, but I put to you, isn't there limitations there as well? No, I think voice as an interface, the limitation has always been the actual processing and quality and efficiency. And again, I've ranted many times and we're going to be addressing Siri and Apple in just a bit, but I actually have had Amazon Echoes in my house and Alexa is pretty good and we use it all the time and we would. now we've switched over to the Apple HomeKit and HomePods and now use it a little bit less and hopefully that's going to get better but I actually think voice has worked and has been a natural medium or natural platform it just couldn't do all those multi-step things I don't know do you remember the Alexa app store like they're really trying this like they're gonna every media company This is when I was working in publishing, was going to launch a new Alexa experience.

7:02The problem was the computing side and the processing side, not the interface. So I actually think this is going to completely revolutionize the way people interact with their phones. In fact, one of the things I've always found fascinating is like my mom in Bengali has the, which is what my parents speak. I speak a little bit, um, to type, it's a lot easier for her to just dictate to the phone. So and actually my wife in Mandarin Chinese does this as well. They're already dictating all the time for messaging. It just transposes it into text. So people, I think, are very, very comfortable talking to their phones to a machine and an AI already.

7:42So this is just going to take it to a whole new level. Yeah. And as you're talking, it reminds me that I did have a viral tweet this week where I pulled up the clip of them talking through a chat GPT4O and having it translate English to Italian and Italian to English and just being this like very conversational and pretty accurate translator. Now, of course, people came out and nitpicked and they said, you know, the Italian accent was off or, you know, how are they going to get slang? And it's like, oh, come on. Like, this is very, very impressive stuff. Come on. OK, so so I do think, OK, maybe this form factor is going to is going to work, especially if the technology is this step change like we think it's going to be.

8:20I mean, this idea of like it responding emotionally or with emotion as opposed to this flat. I can help you with that. I am an AI assistant type of, you know, cadence is interesting. The fact that you can interrupt it is big because these things just tend to talk and talk and talk. And if it can actually stop process and then respond in real time, that's huge. And then there's also, and I'm curious what you think about this, there is personality in these things. And OpenAI basically put out this series of videos of its engineers speaking with ChatGPT4.0. And some people have called ChatGPT4.0 as being a flirty voice.

8:56And I agree it's flirty. but I think it extended to the point where it was actually hitting on these engineers. Let's take a listen. I just need to know, do I look presentable, professional? Well, Rocky, you definitely have the I've been coding all night look down, which could actually work in your favor. Maybe just run a hand through your hair or lean into the mad genius vibe. Your enthusiasm is what's really going to shine through. I don't have a lot of time, so I'm just going to throw this on. What do you think?

9:34Rocky, that's quite a statement piece. I mean, you'll definitely stand out. Okay, so what do you think about this? Is this going to help the technology take off? Is it too flirty? It's a little strange, I have to say. All right. I think this is where, and I do not know the gender makeup of the engineering team or the product team at OpenAI, but I would think more women in the room might have helped us because I agree. I actually think this is terrible. I think they should not focus on personality yet. First, focus on delivering experience that is not so dry and mundane and boring and AI voicey that it's unusable or annoying to use, but the interruption in solving that is really interesting because that creates a much more natural flow of conversation.

10:26But creating a voice that's hitting different inflections and trying to speak back to you, then obviously it gets weird because the ideas in the heads of the people who built it clearly are realized in the tone of the voice. And flirty, certainly when I heard it too, and right there, it's a little weird and uncomfortable versus just try to get me the answer that I'm looking for and do the thing that I'm asking in a reasonable way is good enough for now. Ranjan, don't be silly. Of course you want me to have personality. That was the most awkward moment of big technology. And just think about that at scale in AI.

11:11You've just killed off GPT-4 Omni. Well, finally, you have to take a swing that hit. So the one thing that is sort of being missed, I think there's a lot of talk about the use cases, which is appropriate, and talked about the personality of this bot. The one thing I think is being missed largely in the popular conversation is that this thing is much cheaper. And we're going to talk about this with Aaron on Wednesday. But basically, the opening I made the bot, first of all, free to everybody. So everybody is eventually going to get this 4.0 chat GPT. You won't have to pay the$20 a month for it.

11:44and for developers they made it 50 % less than the current chat gpt4 to use and i think this is interesting because it shows kind of a key strategy i'm thinking about writing this you know for my newsletter or maybe just putting this as one of a series of takeaways but open ai is taking this massive swing right it's really like get it when it's in get big fast mode and i think that was kind of missed here which is that they are making a big swing for users even if it's going to cut into their profits or even make them unprofitable. By making this available to everybody, I think they know that ChatGPT flatlined and they need to kick it up a gear.

12:20And so that's why we're seeing ChatGPT 4.0 free for everyone. And then for developers, and Aaron's going to get at this, they want to make this technology so value efficient for developers that they almost can't help but put it in their products. I think it's a very interesting swing. I don't think it's being fully appreciated, this strategy, but it is a strategy shift for them, or at least, you know, this very clear strategic direction. So I'm curious, you know, as the margins guy, what do you think about their decision to effectively forego margins for size? I think this is a confused decision.

12:55And actually, I don't think it's a good one. And I will say for the upcoming Aaron Levy episode, the entire conversation around the economics of AI was the most fascinating part of it, like really connecting the technology to the business and trying to think about what does this mean for startups or big tech companies. And I think this is a confused effort by OpenAI because on one side, their revenue stream around premium subscriptions and like at that consumer facing level has been the primary driver of revenue, at least from what we understand. We don't have a clear visibility because they're a private company.

13:36and giving it away for free will increase users, but then propensity to pay, is that really going to convert at any point at the consumer level? I don't know. And then on the other side, bringing down the cost 50 % is actually amazing. And I am personally excited to start developing even more and not having to worry about surprise bills showing up. But then you start moving towards Facebook, or sorry, Meta and Llama 3 and open source efforts where all you're paying for is the compute, you're never going to compete against that because that is free. So in a price war, I don't, I mean, the open source is always going to win.

14:16And again, so this is very exciting overall for developers everywhere, for users everywhere. I don't think this is a good move for their business. I mean, maybe it's something that they didn't even have a say in. Like maybe it's just inevitably they know that price of this stuff is going down to zero. And I just wonder what it means for what OpenAI becomes. If that is going to happen, then actually going back to the idea for Omni and what was unique about it, then maybe all the value is accreted at the user interface layer. And the winner is going to be the one that has the best user experiences, which is kind of what happened on mobile as well.

14:56So if you think about it, maybe, okay, that would be the positive take from OpenAI's perspective. If they're showing that they're recognizing the actual model value, like paying to call an API to a large language model is going towards zero, that they're going to have the coolest voice and visual and video experiences on top of that. And that's where they're going to eventually make money. Could be. It's a generous interpretation. It could be good for them. But I don't know. It's going to, this whole conversation has made it real, you know, clear that it's going to get harder and harder for these companies to really prove how they're going to make money, especially the AI only ones like OpenAI versus a Google or Microsoft where this is just an add on.

15:43Right. And then the other side of this is that you and Aaron will address this next week is that basically like if intelligence, artificial intelligence becomes just so pervasive everywhere and it's just coded in to every single bit of computing, then the price doesn't have to be that high to make a lot of money. But there's also other ways that they can make money, and that is potentially licensing the technology to big technology companies. And let's talk about this report, because you have been begging for Siri upgrades. So there's a couple of big pieces of news as Apple gets close to WWDC.

16:20Here's the first one from Bloomberg. Apple nears deal with OpenAI to put ChatGPT on iPhone. And this is a very interesting way that the story is phrased. Apple Inc. has closed in on an agreement with OpenAI to use the startup's technology on the iPhone. The two sides have been finalizing terms for a pack to use ChatGPT features in Apple's iOS 18, the next iPhone operating system. And like when people saw this report, I think they immediately went to, aha, they're putting ChatGPT in Siri. But the report explicitly does not say Siri, right? It says in the operating system. And that is very interesting to me, where maybe the first bit of news that we're going to get from WWDC is that the actual iPhone operating system changes to interact with you from a generative AI perspective before we even get to Siri.

17:12How did you read this news, Ranjan? Yeah, I well, first I read this news and had to almost question my life decisions because how excited I was by it. because as regular, right? Like this was - Friday confession series, here we go. Unfortunately, the single most exciting piece of tech news I have had because regular listeners will know how painful and angry Siri makes me. So that part's good. I think what you just said there, I agree, is where it gets interesting because does this mean there's no more Siri Or does Apple itself allow you at the actual device level to change the default voice interaction?

17:57Like I can select OpenAI. Maybe they start to even allow other voice assistant or LLM driven chat assistants like at the device level. So I choose. Maybe I even have to pay OpenAI via subscription. But basically, they open up the device to allow users to access and have choice around which LLM they're interacting with. And as you said, it could be both when I hold the Siri button and talk to my phone. Maybe it is. And when I'm searching for an app or searching for an answer at the actual search bar level, that becomes LLM driven. So I think I really hope they have something exciting. I have a totally different interpretation of what this means.

18:42And now that might happen, by the way, because we also have this New York Times article saying that they are planning to upgrade Siri. And I think that will happen. If anything, if the Times came to Apple and Apple said, no, we're not going to upgrade Siri, like Apple didn't comment for the article. If they weren't going to upgrade Siri, they would just on off the record be like, your story's wrong, don't run it. And the Times would either not run it or adjust it. But here's what I think their partnership with OpenAI might look like in the operating system. Now, we've heard so much about this agentic type of interface, right, where the AI becomes your agent or the AI becomes your assistant.

19:17And I think what the iPhone is going to do is you're going to swipe. This is a total guess, but you're going to swipe. And this is just that one bit of open AI partnership. You're going to swipe left once you get into your home screen. And instead of all these like little widgets you have on your phone, there's going to be a little rundown in natural language of what's happening in your life. this is what you have on your calendar this is the weather this is the um what you should be thinking about you might be interested in this in the news by the way here are some new episodes on your podcast so like it actually becomes an assistive type technology on the phone using the natural language of an open ai and that's where i think when we talk about the operating system i think that is how it gets baked into the operating system what do you think that okay that is interesting to me because uh do you ever use ios shortcuts or apple shortcuts okay so there was a brief period i tried to use shortcuts and basically these are where you can kind of connect different apps it's almost like zapier if this then that if people remember these uh or use these tools where you connect different apps and different actions trigger other things in in a different app to create a little experience.

20:29Now, like when Alex referred to, when you referred to agentic experiences, the idea that you have these kind of multi-step large language model, AI driven things that are happening. So you can imagine if your phone knows you're taking a flight on Wednesday night and the weather forecast says it's going to rain on Wednesday night, it will somehow in natural language, as you said, tell you, be cautious. I've checked your flight status. It's still on time, but like make sure to bring an umbrella, whatever it is, like all those pieces of information and things and ways you interact with your phone.

21:06It's crazy to think about how much can change and how many cool new ways to just function with the phone can happen if you connect them and have some intelligence behind that. So that is very interesting. If they promise that and don't actually have it ready for a year and don't upgrade Siri, I'll still walk away disappointed but if they actually still promise that and i will i'll find that pretty amazing yeah and what's clear is there is a bond that's developing between open ai and apple i mean open ai released their chat gpt desktop app this week and it's only on apple on mac os to start not on windows despite the fact that microsoft gave them 10 billion dollars to build this stuff and all the all the server processing yeah where open ai fits into this whole big technology landscape is still so confusing because again we don't know their financials we know they're probably have made they have revenue with chat gpt plus and uh their actual like enterprise api services but we have no idea so it almost does feel like they're going to have to ally with one of the players much more closely and it's not going to be Microsoft now.

22:19And also Apple, at least as far as we know, I think there's been some chatter around their own model development, but is the one big technology player that at its core infrastructure does not have AI models and infrastructure that Google Cloud, AWS, Microsoft, Meta, all these companies have consumer facing or enterprise ready AI at the core of what they do. And Apple's the one that doesn't. Right. And this is actually kind of a good time to bring up this listener email that I got. And we had a listener that listened to our iPad segment last week and I had said, oh, the iPad is just thinner. But they brought up this really good point, which is the iPad has this M4 chip and they wrote, you failed to mention that they essentially released the most powerful computer they have to date.

23:08Not only that, they released their most powerful processor to date in a tablet, not on a computer. The iPad is really a beast when it comes to power. I think that's a great point. And I think Apple was going to have like this computing that they can, they can use and sort of put to put to put right on the problem. And that might be with the new chips, it might be with these servers that we're talking about. There's also going to be the opportunity to expand the hardware. And, you know, for instance, the New York Times story about the improved Siri said that a new piece of hardware may inevitably like have more memory to allow a better version of Siri to operate on the phone.

23:44And that's cool. And that could drive an upgrade cycle. My big thing on the upgrade cycle here is that we're probably not going to see an immediate iPhone upgrade cycle right away from Apple, because this is 2017 technology, right? It's all built on the transformers. I got the chat GPT app working on my iPhone. I have the Claude app working on my iPhone. I don't think I need a new iPhone to run Apple's next generation of AI. But over time, there could be like a truly like synced up hardware software relationship inside Apple devices that allows this stuff to sing. I don't know. I think if you, if they introduce some completely different ways and form factors of actually using your phone, and it's only going to be available on the iPhone 17, I think is the next one or 16 um i i'm gonna buy one at least i'm sure yeah i mean they always i guess they have that option to not make it available to current users but that's somewhat user hostile but i mean this is they do the most user friendly and user hostile somehow at the same time company in the world if i need to buy one more dongle to make this ai stuff work i'm gonna freaking throw the phone out the window ai dongles that's gonna be the big uh revenue driver in their services line.

25:00There was a great line in the Times article talking about how they've made this now a tentpole product, a project where they basically have allocated the resources and the top talent in the company to start working on this stuff. Whereas before they hadn't done it in like previous iterations of this style or this iteration of computing, like voice computing in the HomePod. So that's pretty bullish on the Apple front. I am eager to see what's going to happen there. I still don't know if they're going to let me into WWDC. So I'm trying. Come on, we got to get in. Got to get in. Tim, if you're listening.

25:32Yeah. And just on the note of reader feedback, we've gotten a lot of reader feedback over the past couple of weeks. And I just want to say how much I appreciate people writing in to the email address and like actually having a conversation, especially when things haven't been perfect. And again, like I'm learning from you. We were just on with Dwarkesh over on Wednesday and he talked about how he's learning from his listeners and we're learning from you as well. So thank you for doing that. And when you have, you know, it's a piece of constructive feedback, bringing it in via the email versus the rating is always good, although we always appreciate those five star ratings.

26:06So thank you for that. I almost missed this. Ilias at Skever's out, right? So the chief scientist of OpenAI has resigned. He wrote on X, after almost a decade, I have made the decision to leave OpenAI. The company's trajectory has been nothing short of miraculous, and I'm confident that OpenAI will build AGI, Artificial General Intelligence, that is both safe and beneficial. It was an honor and a privilege to have worked together, and I will miss everybody dearly. So long, and thanks for everything. Just kind of a hilarious thing to write in your post. I'm excited for what comes next, a project that is personally meaningful to me, about which I will share details in due time.

26:45And wait, it gets better. right after he left this guy Jan Leike who was also in this super alignment or who was leading this super alignment group inside OpenAI basically to ensure that the AI shares our values he just posted I resigned and he's out too and it led to like this kind of hilarious wave of I resigned from OpenAI memes on Twitter but but it is it is a big deal I would say for two reasons one is because Ilya obviously was has been exceptionally talented and behind multiple breakthroughs. He is the guy that Elon Musk recruited from Google to start this thing in the first place. So it can't be good to lose him, right, at the very least.

27:24And it might hurt a lot. And then secondly, these seem to be the people that were driving the alignment work within the company, which is, again, it's the fastest mover in all of the AI industry. And it is a little bit concerning, I guess, that they're out the door. So how did you look at this news? What do they know? Through all this, when all the Qstar stuff came out, when the initial Sam Altman corporate intrigue was happening. I just want a clear idea. Like if you know something about the potential end of humanity that's concrete, I would hope you would not just tweet, I resigned and you would tell someone and say, maybe we should keep an eye on this.