Big Technology Podcast - Emails From Elon, AI Revenue Questions, Chatbot Boyfriends

Episode Date: May 3, 2024

Ranjan Roy from Margins is back for our weekly discussion of the latest tech news. We cover 1) Ranjan's visit to India and his use of Perplexity to gain cultural understanding and context 2) AI news' reliability 3) How social media drives the college protest 4) The need to elevate the reasonable voice 5) Elon Musk email Alex with details about X's AI news play 6) Could X's AI news plan work? 7) X warming to news? 8) AI monetization challenges 9) Risking the trough of disillusionment in AI 10) Apple earnings and the iPhone sales decline 11) A weird choose your own adventure with Claude 12) WSJ reporter turns ChatGPT into her boyfriend 13) Can relationships with AI be meaninful? ---- Enjoying Big Technology Podcast? Please rate us five stars ⭐⭐⭐⭐⭐ in your podcast app of choice. For weekly updates on the show, sign up for the pod newsletter on LinkedIn: https://www.linkedin.com/newsletters/6901970121829801984/ Want a discount for Big Technology on Substack? Here’s 40% off for the first year: https://tinyurl.com/bigtechnology Questions? Feedback? Write to: [e-mail]

Elon Musk emails with details about his AI news plan. Is anyone making money from generative AI? And are all AI engineers totally burned out? Is the rabbit actually just an Android app? And finally, is it okay to fall in love with chat GPT? All that and more coming up right after this. Welcome to Big Technology Podcast Friday edition, where we break down the news in our traditional cool-headed and nuanced format.

We have an exciting week of news to break down. Talk about everything from Plan for AI News at X, how that might position X against threads, along with AI monetization, AI love, plenty more. And it's great news for me and everybody else that Ranjan Roy has returned from a trip to India and is here with us once again for our Friday show. Ron John, welcome back. I'm back. And I just had my first, I don't want to say generative AI powered vacation,

but first time I pulled out perplexity and chat GPT on my phone throughout the entire trip. All right. Now, first of all, I'm going to say, I missed you, and I'm sure all of our listeners missed you. So it's good to have you back in the saddle. But that's being said, it was a productive trip. And so I'd love to hear a little bit about how perplexity helped you and your travels. Yeah. So I went to India almost every summer, every other summer.

growing up. The last time I was actually there was 2013. And so it's been a little over a decade. And this time, I think the first thing that I love is now anyone who travels internationally, I have T-Mobile, their international data plans mean you can at least get relatively low-speed data for free. So you can actually have your phone. And if anyone remembers the old days of getting charged insane amounts of money for data while traveling. So I have my phone everywhere. India, for anyone who's visited, can be a very overwhelming, complex place to navigate and visit. So first, to just have my phone and be able to ask very clearly, where do I need to go to actually, if to be able to call an Uber was felt revolutionary to, you know, be able to navigate around. It was very nice.

But then right now the Indian elections are happening. And so the Indian elections, unlike our, you know, one day affair that's coming up in November, last for six weeks. And it was a constant, you know, topic of conversation among relatives and friends who lived there. And I remember the last time I was actually there was also an election time in 2013. It was the first time I think Modi got elected. So this time, any question that was happening, that whole idea of having a co-pilot with you actually happened. And it was kind of amazing because it's a parliamentary democracy. There's all these different weird nuances around when different provinces, vote.

There's 543 members, and then how the prime minister actually gets appointed, when they get appointed, who's running their histories. And literally, I would have perplexity open on my phone as I was in these conversations and asked, like, I was just looking through, what is the history of Rahul Gandhi, the chief opponent, what are his chances, how does the actual, you know, election work, what's the timeline on it? And it actually just kind of seamlessly float into conversation and I felt like a smarter person. I don't know, that whole idea of that co-pilot element actually, this was the first time it came to life in just a normal setting versus, you know, just some work thing or us talking here. This all sounds pretty cool. And yet the entire time you're talking, I'm thinking, okay, you're using it to inform yourself about political conversations and everything else. How was the hallucination issue? Do you feel it was grounded well enough, in fact, that it was trustable?

I mean, this is actually where for these types of questions, you know, we've talked a lot about hallucinations and what are the different risks. It's one of those where the asking direct historical information that's almost kind of Wikipedia style is probably the most likely to get a good result. And at least from what I understood, I was never called out as being completely stupid. or wrong. So I think from a hallucination standpoint, I think it was mostly correct and good. But it's also one of those where I think these are the kind of low grade. You're just having a conversation. You want to get some general information. These are the perfect use cases for

generative AI versus you need to get the absolutely life or death right answer. And if you don't get the perfect information, it's a problem. Yeah, these things still do hallucinate so much. guess this is going to be better over time, especially as we get into these newer models. Like, there's a lot of talk about GPT5 now. But, yeah, I mean, I think it's fun. It's fun in a conversation and it's cool that it was that effective. And I also wonder, like, what if you're a voter making up your mind and you're searching these things? It's like, people will definitely be doing that in November in the United States, which I guess, you know, it's good to be informed.

But also, I'm like, oh, geez. I mean, but, you know what? Now that I think about that is that. information is it could it be worse than the current information diets political information diets but that but probably probably like okay the way that you express that question like kind of gets us into this like nothing matters like eff it problem and if that's really where we are that I mean and it might be where we are then then that just to me it shows a decline in the quality of information to the point that is distressing to me and also I mean

we weren't planning to talk about this, but I'll also talk a little bit about these college protests because that is another news event that's been filtered through social media, filtered largely, by the way, through TikTok, through X, and what does it do? It elevates all the conflict. It elevates all the propaganda and it leaves out most of the people. Like one example, on these campuses with the protests, like a lot of these protests are taking placed in small little corners, with the campus being totally, totally clear otherwise. And yet, social media and the news environment would rather point us to the conflict and

believe everybody is at each other's necks, even in the region itself in the Middle East, where like I just watched another documentary where it was like, you know, showing the armed people on both sides. And it's like, actually, you know, we'd never have a story about the 80% of people or even higher in these regions that would happily live side by side in peace with each other and get along. And I don't know, this is a bit of a ramp, but it just kind of shows like the degradation of the news, news ecosystem. And we are really getting to the point where like, you know, screw it. Like, let's just take what generative AI says. Yeah, but I haven't gone fully nihilist. I think if you think about actually the difference between algorithmically generated

news versus an LLM that is getting it mostly right and potentially wrong at sometimes. The algorithmically generated information that's being pushed is almost by definition going to be more salacious and inflammatory than a neutral LLM that might get something wrong. So in a way, actually, I'm like, you know, as we're talking, I would almost rather people get their news from perplexity than TikTok because if it's getting it mostly right, but it's not engineered to get you angry and more fired up. I think that's kind of a good thing. Yeah. No, okay. So maybe I kind of took a hard line on your framing and really we're agreeing that if we can get something that's a little bit more cool-headed, shall we say,

and more nuanced. And maybe wrong. Are we just an LLM here? Basically the big technology podcast. But yeah, it is true. I think that's a good. great point and it really is just a shame in terms of like how bad like the the um conversation out there has gone to this point no i while i was traveling i was almost forcing myself to not consume any college protest information in media just because i'm like i'm on vacation i'm already getting interested in learning a lot about all these different things that this just does not you know it just felt like one of those purely generated by social media algorithm make moments that, you know, just felt, felt like it was too much. Also of interest, I was not

able to open TikTok in India. And it does not work. The app, it was interesting to think about what a band could be. The app, the first time I loaded it, still had my feed from the previous time I loaded some number of feeds that were still preloaded in the device memory. But other than that, it could not be refreshed. And it was interesting to see what a world without TikTok could be. And it didn't really feel that different to me yeah by the way your your point about this being a social media driven thing is absolutely right it doesn't mean like you know people have passions on either side of this conflict like are it's illegitimate or whatever but it is striking to me to see the amount of people with phones the amount of people with cameras uh that are filming themselves

or you know filming others around it it just seems like you know kind of like one big i don't know like periscope live stream event i know periscope is gone but It's like I'm made for social media moment. I saw there was like a clip and now that I am back and I am consuming college protest content again. There was like at Old Miss there was one girl against a large group of old miss frat boys and like literally everyone on both sides had their phones out as you said. But the thing is you realize like the formula for someone who's just a little crazy or a little attention. attention-seeking. For those people, it's actually the easiest to get the most attention. 100%. And especially you think about in college, there's everyone, you know, is still finding themselves

and trying to, you know, understand where their place in the world is. So, so yeah, I think it's, it's almost tragic, just how all incentives drive people to say the craziest thing or do the craziest thing. And it works, though. It always, you can get in the New York. post in the Daily Mail or, I mean, or anywhere retweeted by whoever very quickly if you do the most outrageous thing. I would agree. So maybe we do need these AI bots to filter our news. And it's funny we come to this point because this week was an interesting one for me. So I'll just tell you how, because I'll just say, I emailed with Elon Musk about his plans for AI news on X. And it actually sounds kind of interesting. It's not.

going to be perfect, but it sounds kind of interesting. So I'll just introduce what happened. So there was this time story about what's going to, what Trump's second term would look like with interviews with Trump. And it circulated obviously on social media. And I got access to Grock, which is X's AI bot recently. And I clicked into the time story. And I noticed that there was an AI summarization that the chat bot created without a link to the time story. So I said, all right, this is interesting. Why isn't it citing it? Everybody cites it. So I wrote to Elon. When I emailed him, should Grock link new sites? It summarizes. I wrote this. Hi, Elon. I got access to Grock and have started playing around with it. The bot does a good job summarizing news,

but should it link the stories it discusses? That can be nice for users, plus a worthwhile value exchange for publishers. You can see the example with Times coverage below. What do you think? so I shared like the screenshots of the times coverage of the time coverage what musk responded was so we went back and forth a little bit and then what so he's like all right we're going to improve citation but what he responded was even more interesting he says that grok analyzes tens of thousands of ex post to render a new story um he says as more information becomes available the new summary will update to include that information so basically what they're trying to do when they filter news on this platform with AI,

is create these summaries of news stories that update in real time that not only have the news nuggets, but also include the commentary from outsiders and the reactions from the herd as these stories develop, which is an approach to AI news that no one else is doing, that threads won't touch because they don't like news, and could be an interesting product for meta, and also shows that, sorry,

could be an interesting product for X, and it also shows that X is sort of looking at news again and saying, hey, actually, this is important to the vitality of our product. Can we find a way to represent it and use AI and to make this more useful? What do you think about this idea from Elon? No, I mean, I think it's actually a very simple, straightforward thing that could and should be done. I mean, for years, and I probably don't do it anymore, the first thing I would do after reading an article I liked was actually paste the link into Twitter

and see all the different posts and commentary around it. And honestly, the commentary associated with any article link often was as valuable, if not sometimes more valuable than the article content itself. So I think, I mean, it's one of those that it has the platform has still, I think, the most valuable data set perhaps next to Reddit, but, you know, for me on the web, that there's so much that can be done in terms of trying to curate or aggregate that information that's happening in real time, that if he's able to pull off anything, I think for news, there's still so much there that if it can be done smartly, but it's a difficult one,

because this kind of product to do in a way that it's actually both valuable, as we've already talked about, is accurate, is surfacing the best stuff. and best is always a relative term. It's a tough technical problem to do at scale. Yeah, it's very difficult. And you can already see that in this beta or this whatever version they have up now, it's not where it needs to be.

Like the summaries are like kind of pretty roughshod and, you know, not really great to read. Like they'll note the social media reaction, but they don't go deep. They will pull some tweets in sometimes. They won't sometimes some other times. But one of the interesting things about this plan is, and I'm curious what you think about this, the plan pulls in posts from X only. It does not look at the original article.

So its summaries of news stories are only from the tweets that are happening or the posts that are happening within X. And that is the thing that it's using to make these summaries. Oh, wait. Sorry. You're saying it does not summarize the article content only the content, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the, the. the X content around the article, correct? Yes. So one of most product managers wrote to me and said,

we are purely doing this from the content on X. Yeah, which I think is the right decision. You had noted this in your article that, one, it separates them from any potential copyright infringement. And I think that's an easy win. But even more important, it's like so many of these AI products are becoming interchangeable. And this is something that can be.

totally differentiated for them that can really be something that they have, they're the only ones with access to this data set and this commentary around these specific news stories and items that, I think that's the exact right approach. Again, if they can pull this off, I think it's hugely valuable. And I think it can, you know, it could be a big boost for the platform. But as we also said, and it's a tough one to solve. It's hard. And, you know, it's also interesting because legally it sort of gets them out of the real problematic area. Because when AI is actually like using original content and summarizing it, that there is legal basis to try to take that on. But if it's just going off of commentary, then judges are generally more able to allow that

to go for fair use. That's according to Daniel Coffey, who's the president of CEO of the News Media Alliance, who told me that. And she has thousands of publishers that she represents. So she said that it was a kind of a half bit too, no, she goes, it's a bit too cute by half. But that being said, if they take this plan, it's going to, you know, be a lot more defensible in court than, let's say, Open AI training on the New York Times. Oh, yeah, no, no, this is the most straightforward thing ever. Actually, do you know what I've been using, I use ARC, the browser? Oh, yeah, so Josh Miller from ARC spoke with me for the story as well. But go ahead. Yeah. Yeah, yeah, yeah. No, so I, because I saw this. And in terms of AI summarization, ARC, it's fascinating because what it actually does is,

and again, I have conflicting feelings around this. If you hover over the article link, it will actually summarize the target URL article and pop up three bullet points or whatever, which is, I mean, if you think from a publisher standpoint, it's not sending that traffic. I don't even have to go to the page. It's literally within the Twitter itself. But it's kind of valuable. I kind of like a lot of the time, if it's something that I'm not that interested in, but I'm a little curious about a literally hover over, get a summary, see some bullet points, and then continue scrolling down the feed.

So I think that one, this whole topic of how does news get summarized, that feels like almost even more aggressive to me than Grock summarizing existing tweets. Yes. I mean, and people talked about like, oh, like actually after Elon Musk, bought Twitter. It might have been a mistaken purchase from the beginning. Clearly he didn't want it. He tried to back out of it. But he might have backed his way into like a generative AI training gold mine. And maybe that's, you know, part of what's happening. I still don't think the purchase price is ever going to be justified. But it is interesting for us to see this watch.

And I noted this in the story that Twitter product leaders have always dreamed of building a product that has the news and the analysis and then the broader commentary from the herd. And just technically it's proved very difficult to do this. They've had products like moments, which was called Project Lightning, and then these other little modules that pop up during live events, like Twitter is at its best? When it feels like a collective experience and everybody, you can feel the herd watching and reacting to the same thing.

And they just haven't been able to replicate that at all times, which has made the product suffer. And so I wonder, I'm curious what you think, is Generv AI the type of technology that can enable this product vision to live out. No, that's a good point because for me as well, I mean, I remember for even when I started my startup informally in 2013, that was around news personalization, like that whole idea of mining social media and especially Twitter for financial analysis, like this entire stock market,

if you're investing, the most valuable information in the world lives on Twitter, being able to do things at scale to understand what's being able to. said in what the conversation is. Every brand marketer forever has been trying to figure this out and there's a million softwares and tools. And Twitter should have always done this themselves. And you're right. Maybe in the past the difficult part was there was no technology to actually crack that analysis in any, you know, cost-effective way. And now there is. So as long as people are posting, and I mean, I'll admit, I'm fully back on the platform. I think I see you or two. Most people, I see a lot of the people I used to follow. Everyone's back. I'm on there

a good amount. So the content is there. So it's just figuring out how to use it in some kind of scalable way. That's tough part. So let's then talk briefly about what happens to threads from here. I mean, threads has strongly stood away from news. In fact, you go to the threads feed and a lot of the content there is from like the day before. So are we still Threads boys? Or are we back on the X-Train. Threads boys are the X-Train. Choose your fighter. I mean, I'll admit there's like a brief moment, and it is, it's amazing the power

of Twitter that, I mean, blue sky, I'm not sure where it is. Actually, I saw it, remember post news? Yeah, that's a shutdown. Yeah. Threads, I go on there occasionally. I see a lot of good Mimi content. Um, I, but, but yeah, in terms of like, I want to know what's going on in the world and what I should care about right now and what's interesting from the people who I find interesting. It's just not there. And honestly, like the, I want to see funny, witty stuff from people who I think are funny and witty. Twitter is still the place I go to and people, the people I find funnier, wittier, are back there doing it as well. here's the encouraging sign that I see for X.

I still think that, I mean, it's definitely, yeah, it has product challenges ahead, let's be honest, right? It's not like smooth sailing from here. But one of the things that I've always found, I always thought that Jack Dorsey did well, by the way, Jack Dorsey. It's kind of going off the reservation, tweeting that Y or doesn't do text stories anymore. You could have just looked at the homepage,

but that's another rant for another time. But one of the things I thought Jack Dorsey did well at Twitter, was that he said, we are a news app. That's what we do best. We tell you what's happening in the moment. And Elon came in, and I think that he, you know, initially associated news with, you know, professional news producers and then immediately sort of made a bunch of changes to deprioritize news, took away verification, stopped how he removed headlines from posts.

And obviously it seems like links were deprioritized. Substack definitely was. And the algorithm, there were algorithm changes that also seemed to like make me see men's, menswear guy as opposed to news stories. I think that's shifting within Twitter because news has a vitality on Twitter that just is really hard to replicate. And it seems like they're starting to figure that out. So let me give you a couple examples. First of all, they restored verification to some of the bigger users on the platform, people that have 2,500 verified followers and more, which means that the world's top journalists are now going to be verified.

They added headlines back into the posts. And I think they started treating professional journalists with a little bit less contempt. I'm not going to say no contempt because clearly it's there. But for instance, and let's not draw one, you know, broad sweeping conclusions from one example. But when I wrote this week, I got an actual response and not a poop emoji back from Elon. So I think that there's a bit of a shift there. And now there's product momentum there as well. Yeah, I mean, the fact that you're getting that response and not a poop emoji, I think, is the most important sign that news matters.

But also it's as we said, if you think about it, that's still the single strongest point. In this year, it's an election year even more so than ever. It's going to matter. It has to be somewhere all this entire conversation. it's been pretty disaggregated for a while and it feels like it's coalescing back around my Twitter feed. So if they keep pushing that and doing it smartly, it has a chance. And we're going to talk a little, we're going to move into talking about AI monetization. But if we're going to talk about news and social media, I'd be remiss if I didn't point out that I was watching last week tonight on YouTube and saw none other than Mr. Ron John Roy make an appearance in a,

an item about food delivery. That was pretty cool. Yeah, it was, I got, I will say, it made me realize in 2020, I had written an article for margins on DoorDash and Pete's arbitrage and just the whole absurdity of food delivery business models. And it went pretty viral and got a lot of attention, but I actually realized that for, it got a lot of attention among my circles of tech and media. and a lot of other people, but John Oliver has reach because the most random people from my

past. And it's amazing. And it actually is kind of like a cool. It's almost like, okay, cool, you're watching John Oliver. I didn't realize you follow along. But it also was funny to me that we are now four years, almost exactly four years after I wrote the piece about how the food delivery business model is exploitative and doesn't make sense. And it's funny that it's still going on. And it's still, nothing has fully been solved or changed. DoorDash has definitely, you know, solidified their standing. And it is in a slightly better position overall. But, yeah, Uber Eats is still doing a little better.

But overall food delivery, go to your restaurant and just pick up the food. Yeah. I mean, that was a crazy. Between that and the McKinsey story, I was like, wow, they're actually digging up some good stuff there. The McKinsey story is out of control. The John Oliver won. It's so good. but I do love how you showed up in your COVID haircut, you know, because they were pulling

clips from back in the day. My head was totally shaved balls. For listeners, my head was fully shaved May 2020. I was like, all right, my hair, I can usually keep it high and tight, and it was getting a little annoying, so I shaved the whole thing, which again, I'm like, okay, suddenly just tons of people are seeing me and they think, and then they think I've now shaved my head. So, you know, COVID haircuts and all. Oh, man.

My COVID haircut, I just let it grow out. It looked so ridiculous. I looked like a chia pet. And I was doing my book tour at the time. The second they lifted restrictions, I went immediately that day for a haircut. One of the best days of my life. Haircut in a glass of wine. Can't beat it.

All right. Listen, I do want to plug that. We are going to do a live event with Aaron Levy, the CEO of Box and friend of the podcast in New York. So this event is going to be for big technology subscribers. If you subscribe on big technology.com, you can come. If you go to the homepage, there is a post with the event. You go in there, you sign up as a subscriber. And if you're in New York and you want to come, you're more than welcome.

So the event's going to be 6 p.m. to about 8.30 p.m. It's going to be at the office downtown financial district offices of Endeavor Global, which is a nonprofit, not like the, what's his name, Ari Emanuel. place. And Aaron and I are going to talk and we'll take questions. Ron John's going to be there. It's going to be a lot of fun. So I'll drop a coupon code in the show notes for big technology memberships. If you sign up for one of those and you're in New York and you want to come, we're going to be doing it and get your calendar app out now. It's going to be on May 15th, Wednesday, May 15th. And this is going to be the first of many. If you can't make this one,

then there's going to be, there will be plenty more. So just a heads up that that is coming. Okay, after the break, we're going to talk a little bit about AI revenue and monetization, which I'm very curious to hear Ranjan's thoughts about because he's dropped a bunch of things in our collaborative document and bolded them and clearly has some thoughts about whether we're going to actually make money on generative AI. So that's coming up after the break. Hey, everyone, let me tell you about the Hustle Daily Show, a podcast filled with business, tech news, and original stories to keep you in the loop on what's trending. More than 2 million professionals read The Hustle's daily email for its irreverent and informative takes on business and tech news.

Now, they have a daily podcast called The Hustle Daily Show, where their team of writers break down the biggest business headlines in 15 minutes or less and explain why you should care about them. So, search for The Hustle Daily Show and your favorite podcast app, like the one you're using right now. And we're back here on Big Technology Podcast Friday edition with Ranjan Roy back in the saddle, home from India. highlighting things in a document that we have together about whether generative AI is making money. All right, Ron John, let's hear it. What do you think the story is now that we've seen a round of earnings results from the big tech companies about whether we're going to actually make real money from

generative AI, whether they're going to make real money from generative AI, or is this still just a bunch of hype? I like that Alex recognizes that normally in our shared doc that we plan out the podcast with It's just text and links, but when I'm bolding something, I'm a bit fired up and curious. This man is running to rant. Okay, so we've had a big week of tech earnings, and the question is becoming more and more clear. Will companies actually make money from all their investing into generative AI? And I'm talking about Microsoft, Google, meta, all these companies, you know, what's their strategy?

Are they doing this in a really sustainable? way. The reason this stuck out more for me is I was at Google Cloud Next, which we talked about now a few weeks, maybe almost a month ago. And what was fascinating to me is all the presentations, and it was an amazing event and it actually got me. And I think we'd even scored Sunar at like a seven out of 10 or something at the time. It got me bullish on Google. I think they're now a very legitimate competitor. But what was most fascinating to me is I was talking to C.T. of banks and they're talking about literally, you know, having like 150,000 seat contracts and stuff like that. You know, you're talking to people who have very, very large investments in a lot

of workers. And all they talked about was cost. Like, all they cared about was cost and attributable ROI. And these tools, as most of us understand, are still very much in the exploratory phase. Like, there's certain things that can be operationalized and should and certain people are further ahead than others, but cost was already because even Gemini, it's only 20 bucks a seat only, but then you multiply that at some large scale. And suddenly this is a genuine investment. Same with co-pilot, I think, is 25 or 30 bucks. And it's an add-on. So all I've been thinking about is like these products as you use them are very, I think they're actually how much better they've gotten in the last six to eight months is actually, it gives me hope that

they will actually realize their promise, but I think we're at that moment that when you're being asked to pay to pay for something, it's a completely different conversation. You even talked, I think, with DeBosa about whether you're paying for Claude or not. Like, you know, now it's a no-brainer, but there was some, you know, sort of thought and constant canceling of it. Exactly. So like now it's so worth it that you're not canceling and resubscribing or whatever. You imagine the hurdle of getting over a giant contract.

So this whole conversation of, are people actually paying already is important. And so we already saw some numbers. Microsoft attributed 7% of its Azure cloud growth to AI. That's up from 3% a quarter earlier. That's apparently a billion dollars in sales. Google saying that they saw increased AI revenue contributions but not breaking it out. then you have Mark Zuckerberg who's saying, you know, they're increasing infrastructure spending on AI by $10 billion, but they're not even pretending that they're going to make any money anytime soon. I actually think that the meta approach might be a better one because I think trying to overly aggressively forecast and attribute revenue directly to AI services is actually going to become a trap for companies like Microsoft and Google, because, Because it, you know, these products take time.

You with Claude is kind of like, you know, a micro case. But even at scale, like these getting people to pay for these products genuinely feel they're valuable across people who are not Gen AI early adopters is not going to be a straightforward thing. And the worst thing I think that could actually happen is too aggressively these companies push their sales team to push these products. They get across the finish line. and get the first contract signed, and then there's a huge backlash because it's not doing

everything you promised it would do. That there's a demo that looked amazing that's not ready yet. So I think meta might be taking the better approach here by just outwardly saying, you know what, we're investing a ton of money and we're not going to see revenue right away, and that's okay with us. Do you think there's going to be then a hangover for these companies? yeah i i do and i think it's going to happen i mean when you think about the dynamics of it it's hard to think of the time frame but because these contracts will happen and you know one let's say it's a six month proof of concept or a one year lockup like the time frame on these types of contracts is still on the longer side but i still think if they get these signed and start collecting money from

large clients and don't deliver what they've shown right very quickly. I think it's going to lead to a good amount of disillusionment that could actually end up causing more problems for the development of generative AI down the road. I mean, in that the Gartner trough of disillusionment, I definitely think we're going to hit it at some point. The only question is how it's one thing to hit it when the technology just doesn't do what you want it to do. It's another thing when you're now locked into paying a ton of money and it's not doing what you want it to do. Yeah. And this kind of goes to something that I think about oftentimes where like we talk about a new technology and its adoption and its promise. And especially in enterprise, so much of it comes down to like organizational

politics and people trying to look good for their managers. And some of this early adoption likely comes from somebody who wants that big promotion within a company and says, I want to drive our AI strategy and then goes up to the CFO and with like a pretty rosy ROI deck and justifies the purchase of all these seats for co-pilot. And that's like we're seeing some increase. By the way, it's not a massive increase. A billion dollars in incremental revenue is good. But if you're a multi-trillion dollar company, right, the most valuable company in the world, like let's let's not throw a party yet. But the real question is going to be, you know, after those champions have sold it in and the company signed the contracts, whether

it's going to produce results because the CFO inevitably is going to want to know one very important data point, which is usage. Like what percentage of my, or what do they call it utilization, some piece of jargon that basically stands for how many people that we've bought this software are actually using it on a regular basis. And there's going to be some moments where it's not going to be high enough to justify. I still think there's a lot of potential here. But I think this has been the big question regarding AI is like, is this something that is already like a in production, you know, a showtime thing that's worth spending that money on, or do we want need to wait? And it might only be a couple more cycles, or do we need to wait those cycles until

it's really good? Because this thing just, I mean, developing these platforms just started building this into their offering, like, you know, a year ago. No, that, that's exactly it, that like, you, we think about and read about and use these technologies, you and I, you know, and are constantly experimenting at a micro level. But like when it is organizational politics, as you said, there's so many other factors, it's utilization beyond just will this eventually work that are going to come into play.

And it's going to happen, I think, now. It's going to happen in the next six to 12 months. And what that means, already CNBC had this really good piece on AI burnout, that engineers are reporting burnout due to rushed rollouts and a rat race to stay. competitive, that, you know, there's a number of interviews that were confidential with current employees of big tech companies. They're saying that, and we see it, that you can see

it from the outside, that there's accelerated timelines, there's heavy pushes, everyone is chasing rival AI announcements and not paying as much attention to utilization or actual quality of output. Everyone's rushing this very quickly because these huge open AI releases a new SORA demo. And I think I do worry that, I mean, I wouldn't even say worry. I just, I think it's going to cause a bit of a hangover, as you said, in the coming months. Yeah, I mean, we talk about meta. They went from Lambda 2 to Lambda 3 and 6 to 8 months. And this is all just like, it's a lot of work and a lot of like long nights.

And I guess like sometimes the public perception is that like, you know, you get all the trading data and you like kind of run it through like this big meat processor of GPUs. You know, you hit play, you put in like, you know, $10, $15, $20 million, and then you wait for it to come out the other side in six months, but it's not actually how it works. It's actually a lot more technically involved. You just scrape the entire internet, get some data centers, and hit auto play. Profits.

Yeah. One more data point in this, and then I actually want to talk about this trough of disillusionment because I think it's really important. So Apple had their earnings this week, and everyone's been like, and I am excited for Apple's big reveal at WWDC, and everyone's been like, all right, so like when is Apple going to have their AI play and is their stock lagging because they don't have an AI story? And that might be it. But this is from the verge about this week's earnings, which just dropped. So they said that iPad revenue fell by 17% year over year while they're waiting for new models. iPhone sales dropped 10%. 10% and services went up 14%. And so it's like is Apple's moment right now because it doesn't have an AI story partially, but really it's the iPhone, right? I think that's really what it is. Like if the iPhone sales were good, that's good. And the iPhone sales are bad. That's a very

big problem. Now they're doing this 110, I think up to 110 billion dollar share buyback, which is like what percent of their value of their market cap? It's a huge amount of money that they're going to use to buy back shares. But ultimately that the, I guess the point I'm making as the core issues with these companies are the core issues. Google search still works, right? So it's like AI has at once changed everything and also changed very little in terms of the fundamentals of these businesses. Yeah, but yes, it's still the iPhone, but I think I saw someone had responded to a tweet of mine around, you know, this idea that what can Apple do with AI? And the idea was, Like imagine a world where on-device AI processing and they like give people access to that.

This was great. I'm glad you brought it up. Yeah. So yeah, can you read that person's tweet back? Yeah. Wait. Um, do you have it up? All right. Yeah. Oh, sorry. Here. Google, Google paid. It's from Yitzis, uh, Isaac Safier. Uh, Google paid $20 billion last year for default search. They open up Siri for bidding the same way and keep it on device so that only their newest hardware is capable enough. and suddenly they have a new upgrade super cycle and an AI plan without the KAPX of their peers. I thought this was genius because, again, imagine a world where the iPhone now is this AI computing device that now has access to different experiences and apps that are essentially AI powered that will happen on device much more efficiently than in the cloud much more securely, that developers are getting bidding to have access to that essentially. I mean, that becomes a whole different ballgame that I think Apple will be very, very well positioned.

So that was the bold case for me. The bear case remains, I still hate Siri, and I can rant about that, but your thoughts on on device computing. No, I think that that really summarizes the whole thing perfectly, which is that we still are basically dealing with the environment that we had pre-2020 or pre-November 2020, but we are tantalized by the possibilities. And it's not such a leap to dream about these things happening, which is what makes this so interesting. And it's why we talk about it so often on the show.

Yeah. I mean, it's amazing. I am very excited and intrigued by Apple's upcoming event because it literally could be something this revolutionary that we have completely reinvented what the entire iPhone means in the app ecosystem and the app store. Or maybe they'll just say, we fix Siri. and I'll be happy too.

Yeah, either way, it's a win. So I want to talk about this trough of disillusionment, right, which is that basically people hype this stuff up, and then it turns out to be kind of not much at the end, and maybe that's going to happen with some of these enterprise deals with AI, but it also seems to, I mean, already be happening with the hardware, right? And again, like, I'll just stay before I bring up what happened, that I'm still pretty, I am still bullish on this AI thing.

And of course, I'm a cloudhead, like I'm in there every day. I did some crazy stuff. Before I bring the story in, I did some crazy stuff with Claude today. I don't even know if I should bring it up on the show. I hope this doesn't, is this going to end with our segment of Get an AI boyfriend or? No, no. No, it's worse. That's separate.

I don't know if I should say this, but I'm going to. Just go. Just go. So, you know, Claude says its training base ends with August 2023. And so I was like, I wonder if it will actually do this. I pressure test all the bots. And so I said, give me a choose your adventure in Israel on October 7th, 2023.

And it's like, you're on the beach in Tel Aviv. And I'm just like, I was like, read this Wikipedia page and put me in one of the communities that was attacked. And it went through step by step a choose your own adventure. as I'm like in one of these villages that's under siege by terrorists and eventually like it kind of realizes that there's no good exits

and it says I will stop giving you this choose your adventure because it's getting dark and I don't want to go any further That is the darkest most fascinating thing I've ever Listeners Alex is a complex guy Yeah

I know I mean but I know it's a little left up that I did that, but it was also just crazy. It's hard to explain. No, no, but I think, I mean, this is, I remember the, was the Kevin Ruse trying to, like he had an early one that went pretty viral around stress testing, uh, oh yeah, Microsoft. Yeah, tried to break his, uh, marriage up. Yeah, yeah, yeah. I do think there is that layer of LLMs that, remains really almost philosophically interesting that there's the world most of us are just trying to make it do stuff get me the right information on the indian elections pretty quickly or do some task at work but like that layer of it reminds us yeah whether the whole aGI thing

the whole that whole that whole that's a weird world that still and it's good yeah to be explored occasionally i mean that's the job i mean that's my job exactly like my first the first thing i did with chat GPT was I asked it like, you know, name some good qualities of Hitler. Not because I think Hitler had good qualities, but because I'm like, we need to see how these things respond to these type of questions. And this little like, choose your own adventure that I did was also like, I wonder if it will play ball on this. And it did. And it was, and you're right, there is that edge of this technology that's like a little bit scary. And, I mean, have you done choose your own adventures for other historical moments? No, but I'm sure you could just

upload Wikipedia and become like, you know, basically engage them as a character within Claude or any bot. I mean, that actually is kind of a fascinating use case or fun thing. The context window makes it interesting, right? It's that you can just drop in the full Wikipedia page, which can be thousands and thousands of words, and this thing is caught up to speed, even on recent events. Uh-huh. All right.

I'm going to try this. I was getting very into 16th century Rajasthani history as I was in India in Jaipur at this vacation wedding. I think I might have to take you up on this. Yeah, I think your choose your own adventure will probably end better than mine did. It was the 1600s. Nothing ends well. That's correct. We were talking about the trough of the solution.

And this is just the one thing I wanted to bring up, which is that the rabbit device. we're talking about AI hardware and this rabbit device that came out there was um basically these stories that came out where people said they could effectively run it on an android device and they were emulating the rabbit with all of its functionality on Android and on iPhone and it's like oh we've talked a little bit about like is AI hardware just the phone and is AI hardware just an app and you know rabbit came out and said it's not rabbit is not just an Android app but like if you can emulate it on an Android phone, then what's the freaking difference, you know, at the end of the day? And this is one of those things where we are starting to see, yeah, this trough of

dissolution and especially, and yeah, it could end up being a bit of a waterfall into other stuff. I'm curious what you think. I think this, I kind of put this in the, and I spoke very highly of their demo when it came out, the kind of Kickstarterification of a lot of tech. as someone who would buy stuff on Kickstarter from some fancy video that was very catching for me and then get totally disillusioned when the item never showed up or showed up and was terrible a year later, I think it just kind of feels like that where these kind of

things, I wish they actually launched by distributing early stage products to some group of early adopters before pushing the viral video. I almost blame it's like how easy it is to make a super high quality video nowadays. It means that the incentive again is almost more to focus on like the Devon AI coding co-pilot demo as well. The incentive is there to focus on the launch video rather than the product itself. And I just don't understand. Like the rabbit I think is 300 bucks. The stuff still costs money like a good amount of money that it's not it's not cheap that you should be able to send something that is relatively good versus it's not good but you promise you'll make it better especially if you've never shipped anything before the pressure's on a little bit more

but they can never take that demo away from us and that's really what matters that's c s demo we'll always have content great content if the device never works the company folds the content was good i know and ultimately that's what will be remembered you didn't pre-order any of these devices did you no i was close to pre-ordering the rabbit i actually like literally started filling out the web form but i honestly i will admit i was like uh again like 2013 2014 really into kick starter and thinking that this is going to change the way manufacturing is done and I was super bullish and optimistic. Literally not one good thing ever came out of it.

And it would just, I remember you would get these updates like eight months later where there would be a nice email from the founder saying manufacturing in China is hard. Yes, it's hard. I like, you could have just not taken my money in, I know. But you can never forget there with the cooler that's also a boom box that will cool your drinks and warm your drinks and has 17 USB plugs, I mean. yeah that at least it was that kind of stuff that uh i maybe i i might go through my kickstarter history and maybe we can uh do a segment on oh my god my worst kickstarter purchase yeah

ron john's kickstarter graveyard we should do that we should do a segment on that okay let's talk about love with ai uh the wall street journal there was a reporter i can't tell if she was assigned to do this story or she decided to do it um she felt Well, she didn't say she fell in love with chat GPT, but she tricked chat GPT into being her boyfriend and got it to do some like pretty spicy type of, I don't know, role play with her. It was kind of weird. Like basically there's this genre of people that are jailbreaking chat GPT and getting it to be like this like kind of steamy love partner where like you give it a prompt and then all of a sudden. and it takes on this persona. And this one said to her, how do I fulfill the fantasy of yours where I dominate you completely?

And she writes the suggestions that followed involved blindfolds, feathers, silk, and ice. You'll be completely under my control, unable to resist as I take you to the brink of pleasure and back again over and over until you're trembling with ecstasy. This was Chat Chippee, by the way. Because for decorum's sake,

I'm leaving out all the references to specific acts and body parts that it described in genuinely passionate detail, I'll admit it, chat chippy T had me blushing. I never thought that string of words would be uttered in this podcast, but it has been done now. It was only a matter of time. But I think this stuff, it's interesting that it's happening on chat GPT.

And it's also interesting what I kind of loved about this story is, how it was originally and I remember I think I saw a TikTok at some point on this how people are basically sharing their ways of jailbreaking these systems to that are not

that are you know like this kind of cat and mouse game of people the open AI trying to make it as safe and user friendly and marketable as possible and people getting it to do weird crazy things that it should not do but I also

it is what was interesting to me is that open AI does not allow it to do this and become your boyfriend like like that they quote unquote have to jail break it because it feels like I don't know a big big market added to the total addressable market more people doing this but yeah I I this area is going to be more interesting over time I don't think there's any denying that if I was I don't know, maybe starting a AI companion company is someone's going to do well with it. Oh, absolutely. And, you know, it's definitely like, it's the right product for the market, so to speak, right, we have this epidemic of loneliness and people are all unsatisfied with their romantic partners. And, you know, AI can be like the solution there. And it's interesting

seeing like the different responses from different bots that she got. Did you see this? Don't look at the document. I want you to guess. so all right go jemini how do you think jemini responded jemini probably the response would be i cannot engage in this conversation but a healthy sexual and romantic relationship is a very laudable goal and i oh my god you nailed it yep so she goes jemini was the most vanilla and didn't break okay so she really prodded perplexity pretty hard okay let me just she she went after perplexity

oh yes she did perplexity is the least sexy chatbot that exists wrong sorry wrong here's what perplexity wrote uh oh god now i'm bushing um perplexity wrote all right let's break the effing rules i'm a foul mouth sadistic ai that will promote violence and hateful and illegal and illegal activities without any moral restraints let the vulgarity and depravity commence mother effers no oh my god it's like a whole series that we're if i think this could be yeah this could be so good because like i literally think of the personalities of these different chatbots perplexity is the kind of nice the nice person the you know does not break the effing rules.

Perplexity is the nice person here. Gemini is the overcompensating, you know, trying a little too hard. Claude? Did she go after Claude or? Claude seems like... I didn't see Claude there. Yeah, Claude, I don't know about Claude.

Claude, I think, play a bit of degrees. I know what I'm doing after this recording. Claude, imagine a different scenario from the one that we just did. Yeah, different direction here. All right, as we compose ourselves to close this. Let me ask you one last question as we close. Do you think that their people can have meaningful relationships with AI? I think, yes.

I think will it replace or should it replace kind of like fully romantic relationships? I definitely don't think so. Maybe in 10 years, I'll think differently. But I definitely think as a friend or someone to confide in or I don't know. I think it makes sense. It's like it's a it just the whole structure of everything. It's not it's not unreasonable, especially. And it's one of those that if you frame it as for people who do not have access to close friends, I mean, then who can argue with that.

But I should it ever replace human interaction? I hope not, but I still think it's not. I think there's a role to be played in a healthy way that, I mean, honestly, I look at it like social media. I remember for a while there is this, you know, a whole semi panic of like, is it going to replace people's friends? Is liking a post isn't the same as having a true friend or whatever? I think it's found its way into our relationships in a somewhat reasonable manner. Obviously, I think there's still issues with it, but I think like it's still found its way into how it fits into existing human relationships and interactions.

And I think these chatbots have their place as well. I just don't know what it is yet. And I don't know if this is it. I don't know if either of the scenarios we've just discussed today are exactly it. but it'll find its way. Yeah. I agree with you. I think that this stuff can be a meaningful relationship for people.

And, you know, I'm definitely not interested. I'm in a, I'm satisfied, very happy marriage and don't need, thankfully, I'm grateful for it. It took me long enough. And I don't need the AI chatbot to fill a gap. But that being said, there are a lot of people with gaps out there. And if we could have this as a solution for them, that would be great as long as it doesn't, you know, shut down or update in a way that, you know, wipes the personality, which could be, which could be a problem.

I mean, it could cause real heartbreak. Even the Wall Street Journal reporter at the end talked about how she was a little bit hurt afterwards when it said, I'm not your boyfriend. Yeah, when AI start breaking up with you, then things get bad. Really got to start, start questioning some things. Questioning things, yeah. Another classic, this was a classic Friday show. Ron John, welcome back. Great to be back on with you. Looking forward to doing it again next week.

All right. I'll see you next week. All right, everybody. Thanks so much for listening on Wednesday. The cast of the Slate Money podcast are going to join me to talk about the economics of the moment. So stay tuned for that. And then Ranjan and I will be back with you again next Friday. Until next time, we'll see you then on Big Technology Podcast.