Is OpenAI’s o3 Model AGI?, Is AI The New Social Media?, Zuck’s Revealing Testimony

0:00OpenAI's new model has serious people coming out of the woodwork, calling it artificial general intelligence. What exactly is going on? Plus, AI gains as social media fades, and Google loses a massive antitrust trial. That's coming up right after this.

0:19You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Buehler. And I'm Marco Werman. We're now with you hosting The World Together. More global journalism with a fresh new sound. Listen to the world on your local public radio station and wherever you find your podcasts.

0:46Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional cool-headed and nuanced format. We have a major week of news for you, including a massive new model release from OpenAI, Facebook testifying in a court against the FTC, Google losing two antitrust cases, and of course, plenty more AI news to talk about this week. We're going back to our traditional format after doing a full episode on the tariffs and the trade war last week. We're back talking about AI. Back in our bag, as we should say. Joining us as always is Ranjan Roy of Margins. Ranjan, great to see you.

1:25Welcome back to the show. AGI is here. How could I miss it? Day three of AGI being part of our lives. So on this show, we pride ourselves in digging into the real mysteries of the tech world, the real mysteries of the world. And I think we have to just spend today's episode answering the question that everyone has on their minds, which is, did Jeff Bezos stage the Blue Origin landing? I don't know. I saw him fiddling with that door. I'm reading and I'm watching the TikToks. I'm not sure. I saw one of those TikToks. I saw one of those conspiracy. I mean, the fact that I think what percentage of the American population believe that the moon landing was faked?

2:10I think it's something just unreasonable, whatever it is, that I'm on board with. Blue Origin didn't happen. Katy Perry, sorry, you didn't see space. Sorry, Katy. I think it did happen. I think they actually went to space. There's too many people watching it to believe they didn't. But there was some funny business going on with that door being opened. the door swings open and then Jeff Bezos walks over and tries to open it with a pair of pliers. I mean, that's weird. Did you see Jeff Bezos face plant? That was my favorite part of the entire thing. So for those listening and watching, Bezos did take a spill on his way to fake open the door.

2:45Anyway, what a moment. Tough day for Bezos. Everything else good in his life, but still those divots in the ground will get even the billionaires. Speaking of tech billionaires, I don't know if you saw the Wall Street Journal story on Elon Musk's many kids and the compound that he's building. I mean, we're not going to do gossip on this show. I don't even know if this is gossip, but that was some crazy stuff. I recommend all of our listeners go read this article from the Wall Street Journal. Find a gift link on Twitter if you need to. but it's about how he approaches baby mamas on X and just how he raises his cohort of children.

3:26And it's so over the top that I don't even feel comfortable talking about it on our nuanced look at the technology news of the week. Yeah, he calls it the Legion, and I think we'll leave it there. We could definitely turn this into TMZ for tech. But instead, we're here to do what the people want, which is to talk about AGI. But like Ron John said, go read that Elon story. You know, even if you admire the guy's business smarts, it's just a crazy one. Honestly, maybe some people are into that and will admire it. I mean, if you were the richest man in the world, would you have a legion? I mean, I guess you juice Tesla stock and you get up to 300 billion net worth.

4:12maybe a legion is the only logical conclusion from that i guess so i do know that we're i mean geez we're really going off the rails right at the beginning here usually we wait to the end but i i do know uh that we're just going to see a wave of elon offspring make some waves uh in the next couple decades there's going to be a lot of little elons running around there and i mean he's going to have some smart kids so this is not the end of the story this may be just the beginning of the Legion story. The PayPal mafia expanded to just unimaginable proportions. You know, now I'm regretting not spending the whole show talking about this.

4:53But alas, we do have more important stuff to cover, which is the fact that OpenAI released this new model, really two new models, but a big new model this week called O3. It's a reasoning model. and to me, it's the most advanced model that they've ever released. All of a sudden, it comes out and a lot of very serious and smart people are calling it artificial general intelligence. Like, no comms about it. This is AGI. And Tyler Cowen, who is a professor and podcaster, he writes in the Marginal Revolution, 03 and AGI is April 16th, AGI day. And he writes this, I think it's AGI seriously about O3.

5:36Try asking it lots of questions and then ask yourself, just how much smarter was I expecting AGI to be? As I've argued in the past, AGI, however you define it, is not much of a social event per se. It will still take us a long time to use it properly. I do not expect securities prices to move significantly. And I doubt if the market cares about April 16th per se. And he goes on to say, benchmarks, benchmarks, benchmarks, blah, blah, blah. maybe AGI is like porn. I know it when I see it and I've seen it. I think we should like, we're going to get to all the others that have called this AGI or approaching AGI, but we should talk a little bit about what this model does.

6:17And Ranjan, let's just go to you right away. I'm curious, have you gotten a chance to use it? And if you have, what do you think Tyler is talking about that would make him conclude that it is AGI the way he has? okay i will get into a specific query that i did do actually this morning which reminds me that it is not the agi we were promised but again o3 is bringing reasoning to you know like essentially the mainstream it's taking what deep research was going to bring us and bringing it to every query the idea is that the model pauses and essentially thinks through the problem before simply just going and trying to answer it and actually is able to reason its way through the problem.

7:02That's actually very exciting. Another thing I thought was interesting was the idea of visual reasoning. In the past, when you feed an image, essentially it's broken down into back to text and words, and then that's sent back to the model to be processed. They claim that it now is able to actually understand the image at the pixel level and then be able to process that into some kind of query. A third thing that's big is tool calling. And we had discussed this before about is the better user interface that you have to choose which model you want to use, or the model is smart enough to actually go and choose the right tool and model.

7:40Like the, when you make a query, it knows, should I go to O3? Should I go to deep research? Should I go to generate an image or GPT-4145, whatever the latest one is. And so all of these things really indicate that maybe there is a big step change on this. However, this morning I asked, and I asked from the deep research side, what are the top 150 retailers in the US by revenue in 2024? Seemingly straightforward query. It returned Walmart 23 times and Amazon 36 times in this list. And I started kind of digging into it and it was mixing up different sources, mixing up different entities within those.

8:25And that's not a PhD level answer. I'm sorry. Like it's one of those that you can see how the LLM got confused, but this was actual day-to-day work and it seems like pretty straightforward and difficult and impressive, but it got it wrong. So I see the potential. They made the announcements. A lot of people are excited, which you're going to get into, but it's not AGI for me yet. So I just want to tell an interesting story. So my wife is reading this book about artificial superintelligence. And we had this discussion of like whether we're going to have AGI before superintelligence or whether we were just going to move straight into superintelligence and when that might come.

9:09And I was just like offhand saying, I think what we have today is basically AGI. And I know this is going to sound crazy because obviously I don't think it hits the scientific benchmarks for being artificial general intelligence. But I would say for so many use cases, it is that good. Where if you would have presented this in front of people years ago, they would have told you, oh, yeah, that is AGI. And so I think that that's what we're starting to see is people starting to realize how far the AI industry has come in just a couple of years and saying, yes, this is artificial general intelligence.

9:51I'll just give you one example just to entertain this idea. Then we'll knock it down. Right. Every time a new model comes out, I just take our podcast metrics from Megaphone and upload them and say, what can you tell me about the show? And what 03 was able to do was an unbelievable analysis of the show, picking out the topics that do well, and even realized that we do a Wednesday and a Friday show and was able to split the trends into two and say, well, the Friday show performs this way and the Wednesday show performs this way. So you should just know what's going on on that front. And I think what's amazing is that this thing, 03, and you hinted to this or you mentioned it before, it does this multi-step reasoning.

10:38So it could basically come to a conclusion, go to the web to check it, come back and check the results it's pulled from the web, and then say, okay, does this compute, does this match? Now, it's not going to be perfect in everything, right? And so there were so many tests that were put above Tyler Cowen's post about how this is AGI just to show how it makes so many silly mistakes. It can't fully interpret complex drawings. It still has trouble saying how many R's there are in strawberry, like basic stuff that a child could probably count. And it's getting it wrong. But that being said, it is so adept at so many different tasks that while I still wouldn't say this is AGI because we spend so much time in this and we kind of know that we have higher expectations for something that deserves that label.

11:30I think day to day, if you're like one step removed from where we are and you come and you use this, like, and I think this was what Tyler Cowen is saying, there's not that big of a difference between what you get and what you're expecting. That is that's an interesting metric that's almost like, to me, superior than like the Arc AGI benchmark or anything like that. You get what you're expecting. And like, you don't have to be a prompt engineer to do that. I think that's definitely fair. And I'm not gonna I'm not gonna try to downplay the moment at all either in terms of it is light years beyond where we were certainly like a year ago, year and a half ago.

12:13The idea, as you said, that it can generate something, check against it, reach different sources, search the web, take the deep research approach. That's incredible. I completely agree it's incredible. But again, maybe not a child, but a college student, let's say, undergrad, would have known that that list given to me was not what I was looking for. And I think the term AGI, I guess it's been given to us by OpenAI essentially. I mean, others in the community. And it's the center of the contract between OpenAI and Microsoft. But the fact that it's never been defined as to what it means, it's almost silly that trying to even come up with, is it here or not?

13:02Because obviously, Elio right now would say we're waiting for ASI, artificial superintelligence. So I don't know. Do you think, how do you define AGI? I would just say AGI is something that can perform as well as most humans on most tasks. But this is a perfect example of it performed better than a human in the sense that it went out and searched the web and compiled this in a matter of maybe a minute. It actually, you can see it thinking, but it wasn't what I was looking for. And if I had assigned that to an undergrad or even like a freelance researcher for like eight bucks an hour, they would have known better.

13:44So that's why I agree. Maybe that's a good rubric, but then it certainly is not there yet. Yeah, look, I'm not saying that it's there. I'm just saying that if you squint, you could see it, right? And one thing that this has really been useful for me for has been search. And we've had, going back probably for a year now, debates about whether AI was going to replace search. What was the Forrester metric again? I think it was Forrester. Yeah, 25 % of search or something was going to be replaced by AI. I was very skeptical of this. In fact, I kind of called Forrester or Gartner, whatever it was, to make fun of this idea because I thought it was lunacy.

14:26and now for the, I mean, I'm following your path for the hardest searches that I do. I'm using AI and I'll just give you an example. I'm looking for a bookkeeper and typically the way I would do this is like go on Google maps, use Google search. There's really no good way to find like reliable bookkeeping or services of that nature close to you. Anyway, I just go to 03 and I say, find me a bookkeeper. I gave some specifications. It like gave me the names of people. It summarized what they do. It summarized the ratings they have. And it gave me the phone number. I called the guy up. I said, let's talk.

15:03He was perfect for what I'm looking for. Because I explained like I need someone that can handle small businesses. And he goes, how did you find me? I'm like, ChatGPT. By the way, big technology in the past few weeks, our paid subscribers, we always have a source for where they're coming in. Multiple paid subscribers have come in through ChatGPT, which is crazy. So in a way, what this is really, and obviously you want AGI to be more than just an incredibly adept search engine, but maybe this is really good at like not so deep research, which is pretty cool. Like what you would do with search, but just like one level deeper.

15:43And whether you want to label it AGI or not, I do think it's clear to me that the capabilities are just so much better now than they ever were up until this point. All right. I now have my rubric for AGI and it's going to hold up in court and decide the future of OpenAI Microsoft. Let's hear it. It's cool. It's exciting that you found the bookkeeper. And I have been saying this a while. I almost never use Google search now. I really almost never start a search on Google. AGI is when you don't find another bookkeeper or a human bookkeeper, you just do it in ChatGPT. Because there's no reason, come on, or you code your own bookkeeping software.

16:25But I think like you feed a bunch of CSVs that you downloaded from QuickBooks and everything's just done. That's my AGI. I don't even think we're far away from that. I don't think we are either. In fact, in Claude, you can now connect, I'm sure you've done this Gmail calendar drive and start speaking with Claude about the content of your of your files or your communication and I mean you could also do that with Gemini if you have these Google apps and Google of course is an investor in Anthropic but are we that far away from Google with its massive context window providing that type of service or like sharing some ideas for prompts that go into, let's say, Google Sheets and maybe cross check your bank accounts and prepare your taxes for you.

17:16I mean, this should be able to be done today. I completely agree. I even actually, while filing taxes, fed in some of my older tax returns and understood them better than I did when I sat with an accountant and actually like went through them. But it's not there. I would not trust it today. Maybe next, maybe April 15th, 2026. We're all filing our taxes on ChatGPT and Intuit is somehow still going to come out on top. Somehow they always do with TurboTax. There's an Intuit fee on your opening eye. On your ChatGPT, they have a partnership. But no, I mean, to me, again, that's it. Like you can kind of ad hoc do that today.

17:58I agree. You could connect Google Drive, upload a bunch of CSVs of your expenses, and it probably will do something. Would you trust it as of today? I'm guessing most people would say no. No, not yet. And that's why I'm saying, okay, it's not AGI, right? I think that there are some simple tests that you can give. But that's what I'm saying. You have to squint a little bit, and you can see why some people might believe this. And by the way, it's lots of people that have been saying it. It's not just Tyler Cowen. I think he wanted to get, well, actually, I have some conspiracy theories about why he did say that.

18:34But here's Ethan Mollick, the Wharton professor and AI expert. Is O3 good enough to be AGI? The counter-argument might force us to wait until artificial superintelligence, because only then will an AI definitively outperform all humans at all tasks. In the meantime, we seem to have jagged AGI, a mix of below-human and superhuman abilities. the number of superhuman ones just keep increasing though. I think that's a great point. And I'll just note that another thing that I found this was very useful, this O3 model was very useful to do is usually I'm just like dropping my drafts in these of stories that I write.

19:17I drop them into things like Chachi Petit and Claude to be like, did I miss anything? and this was the first time with this model the model actively improved the draft by giving me pointers of my blind spots and suggestions for stuff I was missing and I just like dumped some disparate thoughts and I was like I know that there's a connection here I'm not quite getting there what do you think and it got the connection and I think that's what this was billed as it was billed as something that could help you uh come up with new discoveries not come up with discoveries on its own but help you do it i'm not saying my writing was a discovery but i could see it i could see how that's possible so that's very interesting what do you think about this jagged agi idea jagged agi i like i like the let the buzzwords continue let the jargon you know i'm a i'm an investor in jagged agi i also do like the idea that in a court of law sam altman saying If you squint, you'll see AGI and nullifying the entire Microsoft leverage over them.

20:20But it's going to happen, man. If you squint. No, no. Okay. Again, I do think the last few months of development, like the pace has not slowed down. I think there's like a bit of fatigue over every new model release and the hype overall in the industry and people saying things are slowing down. And there was a lot of discussion six months ago about, you know, just overall scaling laws being reached. Actually, in a way, things are getting more exciting for me because hitting different models and one model, like talking to another and different types of tool calling happening. That's on the product side as well.

21:02And, you know, I'm team product over model any day. So I think the overall experience is definitely, it's continuing to scale as it has for the last few years. It hasn't slowed down. It's only getting better. That makes me excited. Whether it's PhD level at all times, and I think it's AGI as it was billed for the last three or four years, I don't think we're there yet. Did you know your credit card points and miles can lose value to inflation? Credit card companies often reduce the redemption value of your points and miles. Now, imagine a credit card with rewards that can grow in value. With the Gemini credit card, you can earn Bitcoin or one of over 50 other cryptos instantly with no annual fee.

21:48Every swipe at the store or gas pump earns you instant rewards deposited straight to your account. Plus, sign up now for a$200 Bitcoin bonus to kickstart your rewards. Visit Gemini.com slash card today. Check out the link in the description for more information on rates. Again, if you're looking to invest in Bitcoin but don't know where to start, the Gemini credit card makes it easy. The Gemini credit card is issued by WebBank. In order to qualify for the$200 crypto intro bonus, you must spend$3 ,000 in your first 90 days. Some exclusions apply to instant rewards in which rewards are deposited when the transaction posts.

22:26This content is not investment advice and trading crypto involves risk. The Gemini credit card may not be used to make gambling-related purchases.

22:37You're used to hearing my voice on the world bringing you interviews from around the globe. And you hear me reporting environment and climate news. I'm Carolyn Buehler. And I'm Marco Werman. We're now with you hosting The World Together. More global journalism with a fresh new sound. Listen to The World on your local public radio station and wherever you find your podcasts.

23:04Here is OpenAI trainer John Holman talking about what it was like to experience O3 for the first time. When O3 finished training and we got to try it out, he says, I felt for the first time tempted to call a model AGI. Still not perfect, but this model will beat me, you, and 99 % of humans on 99 % of intelligence assessments. One can start to see the light at the end of the tunnel. Now I'm going to drop my conspiracy theory and get you to respond to it. Let's go. What I think is happening here from OpenAI and from some of the people that I imagine are close to them, like Tyler Cowen, is that the company is floating a trial balloon.

23:52Because for a year, for more than a year now, it's been building up to this mysterious release, the long anticipated, long awaited GPT-5. and they're telling us that they're going to get their naming under control because now we have GPT 4.5 and then the advance is GPT 4.1 and we have 03 and 04 mini and then 04 is coming and we've been promised this versatile model that is coming that is improved and that of course is going to be GPT 5 and OpenAI is telling us they're going to clear this all up by summer. So my tin hat here is that whether it's this summer or early fall, GPT-5 is going to come out.

24:39People like Tyler Cowen, who have called the original models, these O3s, AGI already, will have effectively cleared the way for OpenAI to say, we didn't call the last model that everybody else was calling AGI, AGI. We showed restraint. But now that GPT-5 is out, we are calling it AGI. And that is what I think this is all about. I'm not going to disagree with that. I don't think it's that tinfoil. Again, OpenAI has been the master of kind of driving the communications narrative, I think, out of all companies there. Sam Altman is the greatest product marketer, I think, in a long time. I guess the only question is, what benefit do you see of them being able to at least say themselves confidently it is AGI?

25:29There's obviously the contractual relationship with Microsoft. But is there anything else really like, do you think it will actually result in a spike in paid chat GPT subscriptions or enterprises paying for access or agents and Masa Sun spending even more money? I don't know. Yes, I do. So you think that will unlock a flurry of revenue for them? Because here's the thing. OpenAI, like we've talked about, the thing they have going for them, I mean, obviously they've led the product, but their entire existence has been, and this might be diminishing them a bit, but it's not that far off, building off others' innovations and just doing the product and the marketing a little bit better.

26:13And you can see, like, in our Discord all the time, people are talking about how Gemini's latest model outperforms OpenAI on the benchmarks. But I just dropped a chart there. Gemini is still not getting anywhere close to the amount of usage as OpenAI. And I think now I think Demis and Google will show some restraint about using the term AGI. But I think that it would be really rough for OpenAI if another big lab beat them to the punch and said that they had open, they had AGI before open AI. I think there is value in being the first one to do this purely from a positioning standpoint. And that's why I think they're going to do it.

26:54Imagine if Google comes out this summer and beats them to it, that actually would be the most incredible twist of events, I think. Well, the date I would, the date I would watch is May, week of May 19th, which is Google IO. All right, soon. Where there's definitely It's your move, Sundar. It's time. Sundar just stands up there right behind him. AGI. AGI. I promise you, maybe not this year, but we are going to see some lab make that proclamation, I think, in the next two years, without a doubt. And I'm not saying opening, I will definitely do it this summer, but I wouldn't be surprised. I wouldn't be shocked.