← Big Technology Podcast
Loading summary
Transcript100 lines
- [00:00]
Masa
OpenAI has raised $40 billion, the largest funding round in history. What's it going to do with the money? And Alexa finally debuts. Analysts ask whether AI is a letdown and does it matter that Elon Musk sold X to Xai? That's coming up right after this from LinkedIn News.
- [00:20]
Ranjan Roy
I'm Leah Smart, host of Everyday Better, an award winning podcast dedicated to personal development. Join me every week for captivating stories and research to find more fulfillment in your work and personal life. Listen to Everyday Better on the Linked Podcast Network, Apple Podcasts or wherever you get your podcasts.
- [00:36]
Leah Smart
Did you know that small and medium businesses make up 98% of the global economy, but most B2B marketers still treat them as a one size fits all. LinkedIn's Meet the SMB report reveals why that's a missed opportunity and how you can reach these fast moving decision makers effectively. Learn more@LinkedIn.com Meet the SMB welcome to.
- [01:04]
Masa
Big Technology Podcast Friday Edition where we break down the news in our traditional cool headed and nuanced format. We're running this week's show on Wednesday and we have a big interview coming Friday. So this week we are swapping our shows. We have so much to talk about this week. We're going to talk about OpenAI's fundraising. We're going to talk about the incredible momentum that OpenAI and other AI companies have right now and how AI is picking up speed in a way that it hadn't, at least all through last year. We're also going to cover whether AI still is underwhelming despite the consumer use. It sounds like a bit of a contradiction, but we're going to get into it. And then of course we're going to talk about Elon Musk selling X to xai. Joining us as always for our Friday show, but today on Wednesday is Ranjan Roy of Margins. Ranjan, great to see you. Welcome back.
- [01:50]
Ranjan Roy
It's good to be back. Thank you Masa, sun and Sam for giving me this beautiful news to start.
- [01:57]
Masa
The week with exactly now, Ron John, there's going to be some confused users. They're going to be like this is the Friday show and you just said, as always, Ron John hasn't been on for the last three weeks and we, I know we have new listeners who've come on. So just to go through what we do here, we usually do a flagship interview on Wednesday and then Ranjan and I break down the news every Friday. We'll have our flagship interview on Friday. Ranjan's Bat Ranjan was on vacation. He is back. So you will hear from him consistently on Fridays as we break down the news. And so let's get right to it. Big funding news. And that is an understatement. OpenAI has finalized $40 billion in funding at a $300 billion valuation. That is according to Bloomberg. Ranjan, this is the biggest fundraising in history by far. Last year, OpenAI had the title of the biggest fundraising ever. It was $6.6 billion. This year they've multiplied that by about five times to 40 or even more, more than five times to $40 billion. In normal times, this would be the biggest business story of the year. It seems like it's gone by in kind of like a ho hum fashion where there aren't many people that have been talking about just how crazy this is. I mean, it's right rewriting the rules for private company financing. What do you make of the financing? I know that you're already seizing onto the fact that SoftBank is in the lead. And what do you think OpenAI is going to do with all that cash?
- [03:25]
Ranjan Roy
All right, so first, in terms of SoftBank being involved and is this big news? I think the $40 billion headline number we have to take with a grain of salt because when you start digging into the numbers underlying it, there's around 10 billion that's actually supposed to be raised, which still is a shocking number and is bigger than 6.6 billion from last year. The other 30 billion is, is supposed to be through the end of 2025 and thereafter, but basically be going to the Stargate project, which is a series of data centers that's involving Oracle and others and helping build those data centers. So the idea that this is $40 billion going to allow us to create more studio Ghibli images isn't exactly accurate. Like it's, it's a large number, but it's so convoluted, like so much of these stories, so many of these stories that I think it's not, it's not as shocking as, as it seems at first glance.
- [04:29]
Masa
Okay, so let's break down the numbers. So SoftBank's going to lead the round. It's going to be 7.5 billion right away. So that alone is the largest VC round in history. And 2.5 billion from an investor syndicate that also includes Microsoft CO2, Altimeter Capital and Thrive Capital, which led the last round. Then there's a second tranche of 30 billion that's supposed to be invested by the end of 2025, including 22.5 billion from SoftBank and 7.5 billion from a syndicate. I guess the caveat here is that if open. This is From Bloomberg. If OpenAI's restructuring isn't completed by the end of the year, SoftBank would have the option to reduce its total contribution to 20 billion from 30 billion. So OpenAI really does need to complete this for profit restructuring. That being said, I don't understand how this is not real, Ron. John. I mean, it is the agreement. It's supposed to come next year. Why are we already saying that this is fake?
- [05:29]
Ranjan Roy
I'm not saying it's fake. I'm not saying it's fake at all. I'm saying it's taking a bunch of different announcements and kind of mixing them together. Again, Stargate, we heard numbers as crazy as 500 billion, $100 billion this year. If we remember the announcement on the, in the White House a few months ago. So it's taking part of that and then kind of again mixing it into this announcement where again, It's a, it's 7.5 and 2.5, it's $10 billion coming up front. That is a lot of cash. And we can definitely get into how they're going to use the money and whether that's going to be enough for them to actually handle their burn over the next three years. Because they're supposed to turn profitable. If you remember the numbers in 2028 after burning $7 billion in 2027. So. So they're still forecasting a lot of burn. But I still, I don't know, I think, and I'm not giving the overall press and market so much credit that they're getting into the nuance and recognizing that's why it's not that exciting a story. I just think these numbers get so big and the deals get so convoluted that to try to make sense of it and process it and get excited about it.
- [06:41]
Masa
Yeah. And it's also, I think the main reason why people have been like, okay, whatever, is because OpenAI already announced this 500 billion dollar project, Stargate. And so that people see 40 billion, which is like by far the biggest VC run funding round in history, multiples of what that typically would be. And they're just like, okay, all right, it's 110 of what you told us. And maybe there is a backlash in some ways to these, these big boasts of these large dollar amounts. When you kind of come in and what in any other environment would be the, one of the most impressive financial announcements in history. People are just like, well, it's much smaller. Even your valuation is smaller than the amount that you pledged to raise.
- [07:25]
Ranjan Roy
Well, also when we talk about that valuation, their revenue projections are incredible. Incredible. So they're on Track to make $3.7 billion this year. They're forecasting to triple that next year to $12.5 billion. So first thing, the valuation. And this is kind of what blows my mind, the $2.5 billion syndicate of CO2 altimeter thrive, they're investing at a hundred x revenue multiple. Like we're not talking in the 20s and 40s here anymore. They're going in at 100x on the forecast that SoftBank's going to end up putting in another 30 billion anyways. Like how you even start to get to those kind of numbers is shocking. But then my favorite part about the revenue projection is a third of the. So they're expecting next year 12 and a half billion. That's going to then go to $28 billion. And, and a third of that revenue is going to come from SoftBank. SoftBank spending on OpenAI for all of its own companies and portfolio companies. So you just start to see, I mean the, the mathematical gymnastics and financial gymnastics involved here are, are only worthy again of Masasan. Yeah.
- [08:46]
Masa
In tech circles, I think it's fashionable to say that if you invest in OpenAI, you, you're betting that it's either going to infinity or zero basically that they invent AGI and it is the or super intelligence really. It does everything for humanity. And your $7.5 billion that you put in in 2025, SoftBank becomes one of the biggest bargains of all time. The other side of that is they do just burn so much money trying to serve Studio Ghibli images and you flush that money down the toilet basically trying to enable that behavior.
- [09:21]
Ranjan Roy
Well, that's why the growth number, the user growth is spectacular. Sam Altman has been on a tweeting tear this week kind of talking about biblical demand for the platform, saying that I think that they had added million users over a few months and then added a million users in an hour. And again, the Studio Ghibli, which you had a great show with Brian last week and explaining it's the whole viral format of creating images in the style of a Japanese anime type thing that just went completely viral. I will say plenty of non AI normie friends I saw posting those on Instagram. Even not just on X, like it was real. People, people are going on it, people are doing it. But that's not. That's ex. That's cool. That's not making you money. Like, that's, that's not necessarily the best forecast for you getting to your $28 billion revenue in just three years and somehow also being profitable on that revenue.
- [10:27]
Masa
Yeah. The studio Ghibli thing continued to be crazy over the weekend. My brother, who has not ever sent me anything, AI drops the family chat, flooded the family chat with the images of him and his wife and his daughter, and created multiple ChatGPT accounts to be able to make more images. That is how insane the demand was. And you're right, it's not a profitable use case. Even though we know that OpenAI had lots of signups. I mean, adding a million users in an hour is impressive, but they're melting these GPUs, which cost 20 to 40,000 a pop. So let's just take this to its logical conclusion. You are OpenAI, you just raised 40 billion. Or if you're in Ron John's camp, at least 10. And we'll see what happens next. What does that money go towards?
- [11:18]
Ranjan Roy
This is where, I mean, obviously, and Sam was tweeting about like, if anyone has 100k GPUs, send them our way. We're, we're, as you said, melting on demand. That there we need anything we can get. This is getting crazy. So obviously it's going. They're still positioning this as it goes to the compute, it goes to just. And building large new foundation models GPT5 one day. So it seems like, and we've talked about this a lot, the plan over the next few years is still the same strategy that's been there for the last two launch some new really cool products. But then the entire bet is on the transformational foundation model, GPT 5, 6, whatever. It's going to be, as you said, AGI things that change everything. And that's the only way they're going to make their money.
- [12:14]
Masa
Right. I think it was Dylan Patel who talked about how the next wave of he's going to come on the show from semi analysis. He's going to come on the show in a little bit, in a couple weeks. We recorded already. And basically what he said was the bet is not that they're just going to make a better chatbot, is that they are going to automate effectively full industries, including software engineering. And that is interesting because you're, you're right that this is something we talk about all the time. Product, how the product is important and how it's a consumer and an application company at this point. And OpenAI showed incredible momentum, right? They have 500 million people using ChatGPT every week. That's brand new numbers that they announced in their fundraising. And that would be amazing if there was. It is amazing. But it would be even more incredible if there was no hardware cost to be able to serve the product. Like when Instagram hits 500 million active users. That's incredible. And the best thing about that is you can serve that product without investing billions in GPUs. But OpenAI is investing billions in GPUs, and it's gonna burn that 7 billion. What did you say? In 2026 or 2027? And so I'm wondering, how do we. In 2027. So I'm wondering, Rajan, how do we think about this? Because you at once have this massive investment in GPUs and a very successful consumer product. And I'm trying to make sense of it because I want to say I'm bullish on OpenAI because of all the users. But the bearish thing is the expectations now, after the money coming in, are beyond through the roof. There's no more roof anymore. They're through the stratosphere and it's going to be very, very difficult to meet those expectations.
- [14:02]
Ranjan Roy
No, I think you said it. It's infinity or zero, which is not really the bet and the decision calculus you want to hear as an investor, I feel. But apparently many of the world's leading investors are very happy to hear that. I think that's exactly it. It's that this is not traditional software. These are not 70 to 80% margins. It's not, you know, like no marginal cost to serve new users or almost zero marginal cost. This is a completely different. It's almost an industrial product in the way it's built right now. And they're continuing to go in that direction. But again, in their credit. I will say there. And we're going to get into the conversation around is AI a bit mid right now and is it. Is it creating a letdown? They're creating these moments. They are the household name. They are creating very cool products. They continue to. And again, operator, not great, but looks really cool. Deep research, both looks and acts really cool and is great. Like they're still leading the way on generating excitement across the entire industry. So they have that going for them. But still the numbers are tough. Numbers are tough.
- [15:26]
Masa
So now I'm going to contradict myself on the infinity or zero thesis. And it is interesting because we're always talking about this question on the show about whether the business is going to work. And of course we talk about it, because if the business works, then the products will continue to get better. If the business falls apart, we won't see any more advancement. Everything is linked in this question about whether there's going to be a return for OpenAI, its investors, et cetera. And let me now make the middle case of saying it won't be infinity and it won't be zero, and that is that if you take the applications that we have today, ChatGPT, the image generation, the video generation, the voice generation, all this stuff is quite useful. And in fact, we shouldn't gloss over the fact that OpenAI in a year has gone from 100 million ChatGPT users to to 500 million ChatGPT users. I mean, that's extraordinary growth. And at the pace they're going with these image generation rollouts and every new product thing they did, we know voice played into this, we did a full episode on that. They're they're going to get there, will get to a billion users of ChatGPT. No doubt in my mind that's coming. And what happens is every time there's an advance, it cost a tremendous amount of money to build that advance. But every single time, we also see that companies figure out how to deliver that more efficiently. So I would say if OpenAI stopped development today and just found a way to make what's in chatgpt more efficient, they could run a profitable business with that 500 million or that billion users. And that's how you get to somewhere where a company can persist and can deliver a lot of value and can be profitable, but just has to give up on some of these wild ambitions if for whatever reason they find out, maybe like Yann Lecun said on the show a little while ago, that you're not going to be able to just scale up these models and get AGI.
- [17:24]
Ranjan Roy
Well, okay, to push back a little on that, it sounds like, and we've seen this in all different types of the companies that flamed out. The story was always acquire users at an expensive cost and then you could just turn down your marketing spend and then become profitable. And that did not work for a lot of companies. In this case, the thesis that, like acquire the users and then make the compute more efficient, which I don't argue is going to happen. We saw it with the deep SEQ effect itself that it should be getting more efficient and cheaper. I think the only problem here is that's not the philosophy of how they're building, it's still bigger and more expensive and they're, they're, they've really laid it out that that's how we are going to win. So they're the idea that they're going to really move, like if they're really going to automate entire industries, the economics even for the companies who would be automated aren't there given bigger, more expensive models. So I still think it's a tough one.
- [18:32]
Masa
Yeah, if you're Masa, you don't want to hear what I just said. You want to hear that you're going to infinity and that's the only option. And, and let's put a pin on this automating entire industries boast because we will come back to it in a moment. But I do think we should take a moment just to appreciate perhaps how these products are gaining steam, because I do think that there was another side of this. There's two sides of the AI discussion. One is, is it going to be profitable? Two is, is it useful and is anybody going to want to use it? We didn't really have clarity on the second question up until recently. And as I'm putting together some of the stories that we're looking at this week, I'm starting to say, wow, it really is happening for AI. So again, 500 million people use ChatGPT every week. It's not just the number, but it's the velocity to which they got there in March 2024. So a year ago they had 100 million users. They've added 400 million in, in a year, which is crazy. Now, the revenue side of it, this is from the information. ChatGPT revenue surges 30% in just three months. And the company they say has hit 20 million paid subscribers. This is something that OpenAI has disclosed. That's up from 15.5 million at the end of last year. And this is the information it says. It turns out a lot of people are willing to pay for a chatbot that can code right and give personalized health advice and medical diagnoses and cook up detailed financial financial plans, among countless other tasks. The strong growth rate suggests ChatGPT is currently generating at least 415 million in revenue per month, a pace of about 5 billion per year. And that is significant money. So OpenAI is really on the upswing. Now, one more thing. Let's talk about the other bots. It's not just them. This is from TechCrunch. Chat GPT isn't the only chatbot that's gaining users. We See that Google. This is according to Similar Web, Google's Gemini web Traffic grew to 10.9 million average daily visits worldwide in March. That's up 7,4 7.4% month over month, while daily visits to Copilot, that's Microsoft bot, increased to 2.4 million, up 2.1% from February. Similar web also says Anthropic's Claude reached 3.3 million average daily visits in March and Deepseek had 16.5 in this million visits in the same month. So ChatGPT, of course is way more, but you're seeing growth across the board. And to me, this is just a moment. We have to admit it. It's a moment where it really is coming together for AI, maybe punctuated by all these AI images that we're seeing with ChatGPT. Do you agree?
- [21:22]
Ranjan Roy
I completely agree. I will say this in the last few months and I would actually say even thinking from the super bowl, there's still a little bit of skepticism I would hear from everyone. Everyone I know is using some kind of chatbot right now. A lot of people are paying for it. It's become kind of the norm. Oh, I have a ChatGPT plus subscription or a Claude subscription. But like, it feels like more and more people are putting in their budget 20 bucks a month, which we have been doing for a long time, and, and, and choosing one and it just is becoming more and more part of their daily habit. I think Gemini seems interesting because again, it's free, it's really good, it's getting better, and it's integrated into the entire Google suite of products. So it's still in terms of who could be number two or even overtake number one. But overall, I do agree we, we've passed the inflection point. It's normal, it's. And when people. Again, I was just traveling for a few weeks and the number of people around me I saw taking pictures and putting them into ChatGPT for translations or just even asking questions around travel. I myself, this was an all perplexity. ChatGPT, even Meta Ray Ban asking questions on the fly trip. For me, the days of just pure Google search are long gone. So it's happening. I completely agree with that. Who's going to make money and how? That's. That's a separate question. Right.
- [22:55]
Masa
And this is definitely going from a moment where it goes from toy to being practical and just thinking about the image gen that we saw from ChatGPT over the past week. Yes, it's been fun to turn ourselves into the Muppets but it actually has like a real business use case as well. And this is again from the information there is this company called Sola Wood Flowers. It's a Utah based e commerce company that sells replica flowers mostly used in weddings. It canceled its plans to spend between 150,000 and 200,000 on photography this year after its manager saw ChatGPT's ability to place customers real image into an AI generated scene. So we're going to see this really have I think significant economic impact and consequences in the advertising industry. It's also going to definitely be used in the interior planning, interior design industry. You can. There was a Dallas based real estate developer that posted on X, this was a great, that great thread images of an empty apartment and then asked ChatGPT to show it furnished and that inspiration is, is and it looked good and we're going to see inspiration there. People will design websites and apps. I've seen some of those come through. People will design merchandise with it. You might even see building renderings and custom charts and I thought that was really interesting. Ethan Malik from One Useful Thing, the Wharton professor One Useful Thing on substack he had the new ChatGPT image generator create some pretty good infographics. And his point here was that basically the image is coming from the model, not being sent off to some image generator. And so you're starting to see just really intelligent AI image generation. We can actually get accurate text and real infographics coming into the model. So again, just going from effectively toy to something with real business value.
- [24:55]
Ranjan Roy
Yeah, I think the visual side of things or the image side of things makes this even more promising to reach a wider audience because it just feels and looks more real. Like it's one thing with text it's just not as, I don't know, exciting or enticing. But it's also like when you prompt, understanding what that output is is more difficult to understand the nuance. But to see, okay, wait, if I prompt this one way, this is the type of actual visual representation of that prompt. And if I change a few words that changes the actual style or like really clear images. I think makes this a lot more accessible for people. I will push back a bit that there is this constant gap between what you see as a cool example on X and when you go in the real world and try to create things for your business. And I have been working with a lot of generative AI image, especially for marketing and advertising over the last couple of years and it's gotten a lot better. But to really get it consistent and good enough to push into the marketing sphere. Unless maybe you're a TEEMO ad I think is it's still very difficult to make it like incredibly consistent. It's really easy to be like, okay, create me a fake DTC brand called Ron John's Snacks of Healthy Snacks and it makes like a really nice looking packaging logo and then even a website. I was actually playing around with this and this came out nicely and I'm like holy crap, I could actually turn this into a Shopify website if I actually could manufacture snacks. But anyway, but like, but I'm going to be launching like a thousand random products I think soon. But, but, but, but in reality like to do that on a consistent basis for a real business. It's not there yet. It's getting closer. But when it's why you get this kind of feeling of letdown because you're promised this great thing. You can do a really cool experiment or giblify yourself and that's quick and cool and it works. But when you actually got to go to try to put this in a, in a work context, it's suddenly not as cool like the custom charts. I guarantee you if right now people actually take their data and I use Claude and ChatGPT to actually feed CSVs and create graphs and stuff and it works. But to like upload the charts themselves and to extract data or to transform them into different visualizations. If that stuff's not perfect, it's going to be a problem.
- [27:36]
Masa
Don't fight the revolution, Ron John. Cancel the flower photo shoot. Fire your attitude. I'm keeping, I'm keeping to the moment. It's here.
- [27:44]
Ranjan Roy
Keeping my flower photo shoot. Yeah, at least in 2025. It's in the budget for my replica flowers.
- [27:51]
Masa
I don't use. No, I, I know. All jokes aside, I think you make an excellent point here. We're going to cover it in the next segment. And one more thing about this, I was again listening going through all the examples that I've seen of AI coming through. This one kind of doesn't really, I don't know, I can't say. I hope it's not true because I hope it is true. But I kind of don't feel like this is possibly accurate. There is a study out of Dartmouth again from the information that says that a custom built AI chatbot called TheraBot reduced patient symptoms at a level comparable to traditional therapy. People with depression reported 51% better, feeling 51% better on average. While many people with anxiety reported a 31% average improvement. And this was the first clinical trial on AI therapy via chatbot, according to the researchers. I. I kind of don't even think that last sentence even holds muster. Are you going to go to an AI therapist? Is this. Is. Are we there?
- [28:55]
Ranjan Roy
I. That story actually reminded me of the Evan Ratliff shell game episode, an interview he did where he cloned his. His voice with an AI and sent that to the therapist and sent it to an AI therapist. Like, I don't know. I. I do think this, like, for this, it's chat, it's back and forth conversation. It's relatively structured and programmatic answers. And I say that, I mean, with a grain of salt, but like, it's. A trained psychologist is going to have, like, really structured ways of answering your question, so there's no reason that shouldn't happen here. Whether people are okay with that or feel comfortable with that is another. It's another question. But if it's literally texting back and forth, is it really that different than a sitting in a room with someone and talking to them face to face? It's different, but could it do a relatively good job? I think. I don't think that's crazy.
- [29:53]
Masa
Yeah. I would certainly not sit down with an AI therapist. Not yet.
- [29:57]
Ranjan Roy
At least sit down or text with, like, chat.
- [30:01]
Masa
I wouldn't text with it either. You let these things into your inner self and then they can manipulate you. I just don't feel okay with a computer program doing that. But I do think, going back to the Evan Ratliff episode, it was really funny when the AI therapist is telling his AI bot to breathe into a balloon and fill the balloon with all of his anxiety and worry and then let the balloon float away. And the AI bot is like, I am breathing into the balloon now. The therapist is like, good, good.
- [30:31]
Ranjan Roy
I don't know.
- [30:33]
Masa
Really?
- [30:33]
Ranjan Roy
Is scrolling through your Instagram feed really that different about letting a technical system know all of your innermost desires and then present you with algorithmically generated or curated content to. To answer those needs and desires? Not to get too philosophical here, but.
- [30:55]
Masa
Yes, I think it is. I mean, I really think that you're talking. Instagram is only a subtle manipulation, though. It is a manipulation. I think that when you're. I mean, when you're speaking with an AI therapist, maybe I should try it before I. I knock it, because I have spoken with the replica AI companion, and that certainly opened my eyes once I started talking to that.
- [31:16]
Ranjan Roy
How that go?
- [31:17]
Masa
I. I do. How did that go?
- [31:19]
Ranjan Roy
Yeah.
- [31:20]
Masa
Are You a little too real for me, man.
- [31:21]
Ranjan Roy
Are you in love?
- [31:22]
Masa
Too real? I didn't fall in love, but I could definitely see how people do.
- [31:26]
Ranjan Roy
Y. Yeah.
- [31:27]
Masa
So I do think that, yeah, these things can get real. And what happens if you develop a deep relationship with this AI therapist? The company updates the, you know, the software and the next thing you know, it forgets everything about you. Be a pretty traumatic incident.
- [31:44]
Ranjan Roy
I can just imagine the like scroll bar loading and it's like, sorry, we have lost your data. Please start again.
- [31:51]
Masa
Could you imagine? It's happened before. It's happened with Replica. Okay, one last story about AI, then we're going to move on to. We'll move on to our next segment. OpenAI is planning to release an open weight language model in the coming months. This is from Sam Altman. We are excited to release a powerful new open weight language model with reasoning in the next coming months. We want to talk to developers about how to make it maximally useful. It's the first open weight language model release since GPT2. We've been thinking about this for a long time, but other priorities took precedence. Now it feels important to do. I have so many questions, but I want to hear what you think about this, Ranjan.
- [32:29]
Ranjan Roy
Well, again, I do love. I mean you just raised $40 billion so you can at least make it again, hopefully a year or two without needing to raise again. So you can start to say maybe we will totally open source our models and at least provide the weightings. I think this is one of those odd, interesting things where OpenAI to me still lives as kind of a research house and versus like a fully operational capitalist business. To me there's still this. They want to be a leader among the research community. They want to be a leader among AI thinkers. And obviously deepsea created the entire conversation around the need to or the potential around open weight models, so. But I will also agree, I'm probably equally confused as to why now.
- [33:26]
Masa
I think they're just trying to help them. Yeah, they agree it's the product and it's not the model. So Sam, all about being commoditized and now it's just product that matters.
- [33:38]
Ranjan Roy
But we're gonna invest $10 billion in our next model.
- [33:42]
Masa
That's weird. That is weird. I mean, you're right, it's hard to. It's hard to basically square that circle.
- [33:50]
Ranjan Roy
Yeah, I think there's a lot of circles that are not squared in the overall OpenAI structure and story. But they make good products, they got good models.
- [34:03]
Masa
Here's my Conspiracy.
- [34:05]
Ranjan Roy
Okay, okay, there we go.
- [34:07]
Masa
Elon Musk is trying to stop OpenAI from going for profit, saying that the company abandoned its original open source methods. And what happens? OpenAI raises 10 to 40 billion. We still don't know yet. Predicated on it transitioning to a for profit. What's going to make that for profit transition easier? An open source model.
- [34:29]
Ranjan Roy
I like it. That's it. No, I mean, I genuinely cannot think of another. I tried a second ago. I didn't even feel convinced myself as I was trying to explain my theory. I like that one a lot better. And actually that's going at Elon in that way. That's fun. That's fun.
- [34:48]
Masa
Oh, yeah. I mean, you can just see it now. Your honor, we are open AI. Yeah, just look at this. Listen. Look at all the participation we got from thousands of developers as we requested feedback. Clearly we are serving the market in the way our initial charter intended. We motion for this case to be dismissed.
- [35:09]
Ranjan Roy
Case closed. Masa, your check will be wired immediately.
- [35:15]
Masa
There we go. All right, let's take a quick break. When we come back, we're going to talk about a couple of, I would say, anti AI op eds, including this one from the New York Times. New York Times. The tech fantasy that Powers AI is running on fumes. And another from CNN saying the problem with Apple intelligence isn't Apple, it's AI itself. Do these authors have any merit in their attacks? We will dig into it right after this.
- [35:42]
Leah Smart
Small and medium businesses don't have time to waste and neither do marketers trying to reach them on LinkedIn. More SMB decision makers are actively looking for new solutions to help them grow, whether it's software or financial services. Our Meet the SMB report breaks down how these businesses buy and what really influences their choices. Learn more@LinkedIn.com meet the SMB that's LinkedIn.com meet the SMB.
- [36:15]
Masa
And we're back here on big technology podcast Friday edition running on a Wednesday as we break down the news in anticipation of an interview dropping Friday. Of course, Rhonda and I will be back not this Friday, but a week from Friday. So let's talk about AI being somewhat mid. There is a New York Times op ed called the Tech. The tech fantasy that powers AI is running on fumes. It's by trustee McMillan Cottam. She says, behold the decade of mid tech. This is what I want to say every time someone asks me what about AI? With the breathless anticipation of a boy who thinks this summer he oh my God, I can. I do I have to read this.
- [36:57]
Ranjan Roy
I think you do. I think you do.
- [37:00]
Masa
This is what I want to say every time someone asks me, what about AI? With the breathless anticipation of a boy who thinks this is the summary finally gets to touch a boob. I'm far from a Luddite. It is precisely because I use this technology that I know mid when I see it. She goes on to argue that artificial intelligence is no revolution. It is middling tech. It is something that is promised to do magical things, but when you put it into production, it doesn't think about checkout, she says checkout Automation was supposed to change the experience at the supermarket. She calls it pretty mid. Cashiers are still better at managing the point of sale, she says. Think about facial recognition. That's supposed to get you through security faster. And the TSA's adoption, however, hasn't particularly revolutionized the airport experience or made security screening lines shorter. Artificial intelligence is supposed to be more than radical, more radical than automation. Tech billionaires promise us that workers who can't or won't use AI will be left behind. Politicians promise to make policy that unleashes the power of AI to do something, though many of them aren't sure exactly what. I wanted to hate this op ed, but as I read it, I started to think, you know what? We still haven't seen the killer use cases for AI. Yes, you can use ChatGPT to help you sort through a document a little bit more, but has it lived up to the boast of it, of the technology being that revolutionary and you know, really empowering the people that use it beyond those that don't? I don't know. I couldn't fully hate this story, this.
- [38:43]
Ranjan Roy
Article, this, this story took me on an emotional journey as well because I, I think same thing, I wanted to hate it, but also it raises the point. I feel it's too. It's still selling the promise too short, but it's not lying about the present. And this is where, and I've been ranting about this for a long time, AI companies have a branding problem. And we see this the way Google the Super bowl ads and the way Gemini was presented and the way chatgpt tried to whatever present, whatever they were trying to do with that ad overall, everything, the promise of automating entire industries or these big boastful things or people on X doing threads about 10 crazy use cases I just discovered with the new Claude update, like that is what everyone is building this like the promise around and what they're raising money around. So they have to do that. And that's not where the technology is right now. It's not. If you are very good at using it, you can make it do a lot of those things. And there's more and more again, products being built on top of these models to allow more and more people to do these things. But because the industry is promising that that's the present as opposed to the even near future, it's going to leave a lot of people feeling like this again. My mom going back to me at the super bowl, turning to me and being like, so what can I do with this AI and not being able to give her a clear, simple explanation and these products not delivering that for her. And we're going to talk about app intelligence, our favorite topic, but certainly that as well, not actually delivering on what they show in the ad. So I think this is something the industry and I think this year there's going to be a reckoning with it, that when you are promising too much at a certain point consumer fatigue is going to hit in. It's going to hit and people, you might see those user numbers start dropping and then you can actually hopefully get to the actual work. But I think the industry is moving in the wrong direction on that side and this article captured it.
- [41:00]
Masa
I'm totally on board with you and so much. It's funny because it's the exact emotional journey that I had where I was like, you're wrong. There's so much useful stuff you can do with AI and there were passages in there that felt beyond over the top for me. Here's one. The tech fantasy is running on fumes. We all know it's not going to work, but the fantasy compels risk averse universities and excites financial speculators because it promises the power to control what learning does without paying the cost for how real learning happens. Just this idea of that, okay, we're all gonna, we all know it's not gonna work, just struck me as being totally removed from the details. But then she also talks a little bit about why she is so negative on the technology. And again, it's the delta between the promise and the reality. This is what she writes every day. An Internet ad shows me a way that AI can predict my lecture, transcribe my lecture, while a student presumably does something other than listen, annotate the lecture, anticipate essay prompts, research questions, test questions, and then finally write an assigned paper. How can professors out teach an exponentially generative prediction machine? How can we inculcate academic Values like risk taking, deep reading and honesty when it's this cheap and easy to bypass them. So she's an academic, but I think this point is valid in that the industry has promised to do all this. I mean think about how often we are talking about industry promises of AGI. And I think I'm pretty proud of the fact that on this show we haven't sort of gone with the marketing hype and sort of tried to take a nuanced and cool headed approach to what we're hearing from the industry. And there's a reason for that. We think that it's important for listeners to get the truth here. And by extension, Ranjan, you've been talking for a long time about this branding issue that the promises from the industry that you're going to have, you know, these brilliant AIs that are going to be walking around with you feel like they should be here already. And I think the industry oversells what there is today. It's not there. It's going to take years. Sort of undersells what we already have. Right? There's both this overselling and underselling happen in terms of the actual capabilities and then it's no surprise that you're left with somebody who is not a techie but deals with the technology and kind of looks at you and says, you know what, shut up. Right?
- [43:31]
Ranjan Roy
Yeah, I think that's exactly it that I mean when you. And in that gap between expectation and reality like, like even the ChatGPT generating images for brands, like I worked on something and with the chief marketing officer of a fashion brand and I'm like, okay, here is the product on a completely artificially generated model which this is like a year ago and it's blowing my mind that I've been able to do this. And the first comment is the print on the fabric is not exactly the same and it's a very intricate, detailed print and it's like wait, do you not understand just what just happened? I just created a person and put our, this product on them. And the first reaction was a bit of disappointment because the expectation was it was going to be perfect. And I'm sure like this is happening across the entire industry especially when you get into the more enterprise and professional use cases. And I do think this is, this is the exact branding problem in addition to the fact that she even throws in that like it's been Doge has been an infomercial for AI that the, the use cases and the like where it's, where it's living and who's promoting it is causing some problems too on.
- [44:50]
Masa
The branding side right now. Let's go to this CNN story. CNN says Apple's AI is in a letdown. AI is the letdown. And I think continuing on this Apple beatdown that's been going on in the press for the past couple of weeks, it says the real reason companies are doing this is because Wall street wants them to. Investors have been salivating for an Apple super cycle, a tech upgrade so enticing that consumers will rush to get their hands on the new model. Fact check True. In a rush to please shareholders, Apple has made a rare stumble. The company is owning its error and now delaying the Apple intelligence features to the coming years. And this goes to a little bit. This is actually a very incisive point that this author makes here. In June, they write, Apple floated a compelling scenario in the newfangled Siri. Imagine yourself frazzled and running late for work, simply saying into your phone, hey, Siri, what time does my mom's flight land? And is it at JFK or LaGuardia? In theory, Siri can scan your emails and texts with your mom to give you an answer. That saves you several annoying steps of opening your email to find the flight number, copying it, then pasting it into Google to find the flight status. If it's 100% accurate, it's a fantastic time server. If it's anything less than 100% accurate, it's useless. Because even if there's a 2% chance it's wrong, there's a 2% chance you're stranding your mom at the airport and your mom will be rightly very disappointed. Our moms deserve better.
- [46:20]
Ranjan Roy
Our moms deserve better. I agree.
- [46:22]
Masa
The thing that kills the moms.
- [46:23]
Ranjan Roy
Here's to the moms and picking them up at the airport 100% of the time. The thing that kills me about this is honestly that that query should be answerable at 100% success rate. I'm sorry, Apple, you guys should figure that out. That's a, that is a straightforward thing. But this is again, going back to the problem that I do. Well, actually, I will disagree that Apple's AI is a letdown. And I know regular listeners know that's how I feel. But that is a problem that actually most AI systems and chatbots that have. If you upload a bunch of emails and you ask that exact question into Claude, it will get that answer right. So I think Apple, the biggest letdown, and again going back to the gap between promise and reality, is they essentially promised Everything all at once to everyone. Rather rather than being like, okay, let's solve the go to your inbox and answer all of your travel questions. Travel planning. Make travel planning like a little feature, make it a little app, make it like a pop up Apple Tips with the for Apple Intelligence. But instead the idea was all questions could be answered right away. And that's of course it's going to be a letdown. But again I think AI should be able to today solve a lot of this stuff.
- [47:48]
Masa
Yeah, I wanted to read that out there a Because it sort of harks back to our attempt to use Siri.
- [47:54]
Ranjan Roy
Yes.
- [47:54]
Masa
And failing miserably. And I think you kind of seized on my follow up point here, which is AI can and should get that right. There's no excuse not to get that right. And it is going to start delivering. And that's why like when we talked earlier in the show about how AI is finally hitting its stride, this type of stuff is going to push it even further. I'll give you one example. I was in my Gmail inbox and just used the Gemini. I'm always reticent to use these things because they usually don't work. But I said, okay, I have a pretty complex task and it will be worth wasting the 30 seconds on a Gemini query to see if it can work. And that was I wanted it to pull out all the paid subscribers of big Technology. I wanted to pull out their emails and separate them by commas so I can invite them into our discord. And I typed that into Gemini and lo and behold, Gemini produced the list perfectly accurate from a number of emails going back a month. And I was just able to copy and paste that into the BCC field and invite the subscribers into the discord. That's incredible. I mean it is effectively applying a conversational probabilistic technology into a deterministic scenario and it proving that it can execute and once it starts getting that stuff right and doing it for a broad degree of use cases, whether that's Google or Apple or Amazon or Microsoft or all of them, that's when you're going to see the movement.
- [49:25]
Ranjan Roy
But I think that in that exact example, in that exact example is kind of reference like it's a good reference point to one of the points in the article. She's kind of going at Kevin Roos at the Times and she's like he.
- [49:39]
Masa
Had said it sounds like she listened to a hard fork episode. It was like what?
- [49:43]
Ranjan Roy
Pissed off.
- [49:43]
Masa
Yeah, goes in on Kevin and Casey. Okay, continue.
- [49:48]
Ranjan Roy
Because Kevin said there are people who use AI systems and know that they're not perfect and that those are the regular users, that there's a. There's a right way and a wrong way to query Chatbot and then the authors. This is where we, the people are apparently failing at AI because in addition to being humans with jobs and social lives and laundry to fold and art to make and kids to raise, we should also learn how to tiptoe around the limitations of large language models that may or may not return accurate information to us. I like the line. It's a good line. It's good writing at the end. But also to me, this is where Apple. It's a letdown on their part because they promise to a human with a job in social lives and laundry to fold that they'll get all these right. The example you gave is a perfect example of if you kind of know what's possible and how to ask, it's going to get it right. And it's incredible. But this gap between knowing how to use it, like, that's where either there needs to be more user education or the product needs to get better because the models are good enough to answer all these kind of queries.
- [50:57]
Masa
Yeah, I don't know. I mean, no one is expecting you to use the LLMs to make your life better. Just to me, seems like, all right, tech companies built these tools like go and use it if you want or don't use it if you don't want. Sort of. I don't know why I sent that.
- [51:12]
Ranjan Roy
Comment, but these investors are certainly needing that to happen.
- [51:17]
Masa
They need that, but you don't have to as a consumer. You have agency and 20 million people feel it worthwhile to pay for ChatGPT. So clearly this is working for some people.
- [51:29]
Ranjan Roy
I would argue your agency with Apple Intelligence is. Is a bit limited when it's shoved at. Into every part of the iPhone, end product and accidentally. I call Siri even on my MacBook right now. Yeah, I think they're questioning free will in terms of interacting.
- [51:50]
Masa
Turn that shit off. I don't know. Just turn it. Turn it right off.
- [51:54]
Ranjan Roy
Do you think. Do you know how to.
- [51:57]
Masa
I tried. I know it's possible.
- [51:59]
Ranjan Roy
I haven't tried either, but I might go and do it and we should talk about it next week because I have a feeling it's probably going to be a pain in the ass.
- [52:07]
Masa
All right, next week, folks, you tune in. Ranjan and I will both try to turn off Apple Intelligence live on the air, and we'll see if we can do it. All right, very quickly on this line, Amazon's Alexa is out. And this is from the Washington Post. It's missing some features. The new enabled assistant Alexa is launching on Monday. So that's Monday of this week. But not all the features the company showcased already. Some of the new features that aren't going to be coming out include ordering takeout on grubhub based on a conversation about what you're craving, or using Alexa to visually identify family members and remind them to do specific chores like walking the dog. I guess that's if you have the security camera in the house. Other stuff like brainstorming a gift idea or generating a story to entertain your kids also won't be released until later. So I don't know. We saw the live demo at the release event, but I think this is just another case of Amazon or of a company making a big promise about an AI assistant. Now, at least they shipped it. I guess they shipped something. I try to get it to work on my devices. I have to. I think I have to disable multi language and I can start using them. And so we can report back on that next week. But I don't know, should we be excited that they launched or also just be like, all right, here we go again. They are missing features, of course.
- [53:36]
Ranjan Roy
This is where I am genuinely excited about this. As we talked about a few weeks ago, I might get rid of all my home pods and move back to Alexa. But to me, what was very interesting about that announcement is brainstorming a gift idea is a pretty straightforward generative AI question. Generating a story to enter your entertain your kids. I do that all the time with ChatGPT voice and it's amazing. I'll literally be like, tell me a story about this really specific subject involving my kid. Like this really specific scenario. And he loves it, so it can do that job. So why like using facial recognition to identify family members and remind them to do specific chores? That's a tougher problem. That's. That's a tough problem to solve, but. And ordering on grubhub and not getting that wrong and like at a hundred percent accuracy, otherwise people would be really pissed. I get that stuff taking time, but why advertise that stuff if it's not close to ready? That's just like, guys bring back consumer trust in this. Make people happy, make them excited. You know what? OpenAI is actually doing that pretty well.
- [54:50]
Masa
Like in terms of they are doing it the best of everyone, which explains why they have the product that's working.
- [54:55]
Ranjan Roy
Yep. No, that's why they're at 500 million. That's why people trust it even more because you go there, you actually do you know what I'm going to admit right now, I was unable to giblify an image. I did. I kept hitting a content policy thing when I tried to put a picture of me or my wife up there and like try to gibble fight. Were you able to do people?
- [55:19]
Masa
I have been able to, yeah.
- [55:21]
Ranjan Roy
What is what I'm paying you open AI? I'm paying you. Let me gibble ify myself. I don't know.
- [55:28]
Masa
Yeah. What I would do is just, I would just go to the web version. Have you been using the app? The web version is seemingly more permissive.
- [55:35]
Ranjan Roy
No web version. I actually got a little obsessed with this because I on Sunday I landed back in New York and I all I saw was this all over different social media platforms and I'm like, okay, how can I not do this right now? I'm going to write a whole New York Times op ed on how AI is letting me down because ChatGPT is not letting me giblify myself.
- [55:58]
Masa
It's mid. It's mid. Okay, let's just end this segment. I think we have consensus here which is that we both believe that AI is not going to just fizzle out and it's not a, you know, fake revolution, so to speak, but we also think that the over promising is going to have some serious consequences and we're already starting to see some signs of that backlash.
- [56:20]
Ranjan Roy
Agreed on that.
- [56:22]
Masa
All right, a couple minutes left. Let's just talk about the XAI acquisition. So XAI bought or the X acquisition? Too many X's. XAI bought X, so Elon Musk's AI company bought X. It's kind of a weird deal. So there was one set of advisors working for both sides. It is a value. It put Xai's valuation at 80 billion even though there was no new money in and that increased its valuation from 80 billion to 50 billion. So that 30, that 33 billion that X got is actually probably smaller if you just use the last fundraising amount. We have a professor from UCLA that says it's funny money is telling the Wall Street Journal it's funny money. It's like using Monopoly money to buy Pokemon cards. And as someone who has done that in the past, let me tell you, don't knock it till you rock it. Professor Andrew Verstein. And it is interesting that basically we're seeing AI, which is the next platform swallow social media, which is the last platform, and this is what Axios says. AI eats social media as XAI swallows X. All your X data was going to be used to train these models anyway, and now it definitely is and there's no getting away from it. So that's the headline on the deal. Ranjan, I'm curious what you think about it and if there's anything you think the common narrative might be missing about what this deal means.
- [57:52]
Ranjan Roy
I think it's like using monopoly money to buy Pokemon cards. If that's the common narrative, is the right one. Again, kudos to Elon Musk and the advisors who worked on both sides of the deal because you never see something like that. To raise the valuation of Xai from 50 to 30 to 80 and then simply add in the $33 billion price tag and maybe that's what you're attributing the rise in valuation to be. The simple, you know, add on of the $30 billion or so for X is quite incredible. To then be able to just make up whatever valuation you want for X is incredible because remember, he bought it for 47 billion, I believe it was. They valued it at 45 ish and they said $33 billion with the valuation minus debt. Like basically he's, he's able to just say, oh yeah, it hasn't lost any value, even though we've all seen endless reports. And you can even see it in the advertising when you load your X feed, like just how ridiculous it is that they're not making money, they're losing money, it's not going in the right direction. And he just was able to say, oh yeah, it's worth what it was when I bought it, that's okay. And now it's part of xai, my other company, that has an obscene valuation. It's the same investors on both sides. It's the same bankers and lawyers on both sides. I mean, this one, Masa is jealous of this one. He thought his 40 billion and we're forecasting revenues about my own. The investor actually paying money to the portfolio company. This puts that to shame.
- [59:43]
Masa
Well, the interesting thing is that now Xai Xai's revenue is going to be coming from X, which is interesting. Like X will be the revenue arm of Xai in some ways because you're going to pay for Grok through Twitter or old Twitter, you're going to have ads still coming into X. So that adds an interesting wrinkle to it. But let's end on this. This is the close. Well, it's not fully the closing chapter because we don't know what's going to happen to Xai. But this, let's say the intermediate closing chapter, which makes no sense, but you know what I'm saying. On the X saga, was that a good buy for Elon Musk, strictly from a business standpoint?
- [60:23]
Ranjan Roy
I think it was a great buy for Elon Musk from a business standpoint because he still owns it. No cash changed hands and he got to just label a valuation that he wanted to on a property that is not worth that. So that kind of financial engineering, I think we should all be fascinated, proud and terrified of.
- [60:48]
Masa
Well, I think that says it all. Ranjan, it's so great to have you back. Welcome back to the show.
- [60:53]
Ranjan Roy
It's good to be back. See you next week.
- [60:55]
Masa
All right. See you next week on Friday. Yes. Thanks, everybody, for listening. Special episode coming up this Friday, so stay tuned for that. And then Ron General and I will be back a week from Friday to break down the week's news as usual. We're back, baby. Back in action. Breaking down AI news like it's been no time at all. All right, thanks for listening and we'll see you next time on BIG Technology Podcast.