← Big Technology Podcast
Big Technology Podcast
Is Generative AI Plateauing?, Booming Bluesky, Apple’s Smart Glasses Play
00:57:58
Loading summary
Transcript108 lines
- [00:00]
Alex
Is Generative AI plateauing? As training methods top out, Blue sky is booming as an alternative social network and Apple looks into smart glasses. All that more is coming up on a Big Technology Podcast Friday Edition right after this.
- [00:14]
Michael Kovnat
Hey, I'm Michael Kovnat, host of the Next Big Idea Daily. The show is a masterclass in better living from some of the smartest writers around. Every morning, Monday through Friday, we'll serve up a quick 10 minute lesson on how to strengthen your relationships, supercharge your creativity, boost your productivity and more. Follow the Next Big Idea Daily wherever you get your podcasts.
- [00:38]
Tomer Cohen
I'm Tomer Cohen, LinkedIn's chief product officer. In my new podcast, Building One, I interview some of the best product builders out there, people at the intersection of dreaming and building and learning. Together, you and I will learn from their experiences. If you're just as curious as I am, follow Building One wherever you listen and check out the conversation on LinkedIn.
- [00:59]
Alex
Welcome to Big Technology Podcast Friday Edition, where we break down the news in our traditional cool headed and nuanced format. We have a great show for you today covering everything happening in the world of AI. Very big news. There's concern that training methods that have gotten the generative AI field to here are not going to continue to scale and that's really coming to the fore right now. We're going to talk about that. We're also going to talk about the rise of Blue sky, whether it will be sustainable, and Apple smart glasses play, which is quite interesting. Joining us as always on Fridays to break it all down is Ranjan Roy of Margins. Ranjan, welcome to the show. Great to see you.
- [01:33]
Ranjan Roy
Scaling laws are here, Alex. They've finally come for the industry.
- [01:38]
Alex
I know you're excited about what's going to happen with this, but I know you're even more excited about the Jake Paul Mike Tyson fight. Who you got?
- [01:46]
Ranjan Roy
Neither. Neither. I think Netflix is genius in promoting their live programming by just bringing out two people that no one wants to see win. But I'll still take Tyson if I have to.
- [01:57]
Alex
All right, I'm taking Paul. We can put it on the prediction markets and see what happens. My friend group has been saying that this is an exhibition match basically and a lot of betting sites are looking at it the same way. The folks are saying that it's rigged. Do you think this is rigged or a real fight?
- [02:12]
Ranjan Roy
I think this is a real fight. I don't think Netflix would go into this fully rigged or making it reality tv, but I think it's a good reminder. The blurred lines between reality TV and actual live programming, but I think it's real.
- [02:26]
Alex
I'm going to go rigged. Okay, so we got a lot of.
- [02:31]
Ranjan Roy
Market contracts to set up right now.
- [02:34]
Alex
Absolutely. And now we can talk a little bit about what's happening in the AI world. Where shall we say? There's another fight going on between the purists that believe that large language models will continue to scale if you add more data and compute and power to the mix, and those that say eventually these models are going to hit a wall. There's been a long brewing battle between these two factions, and this week, I think, has been the week where both sides have started to put their positions in the ground and say, you know what? I've won. And that's really come on the back of this great information report that talks about how OpenAI has basically found that it has hit the limit of improvement when it trains with more data, more compute and more power. Here's from the story. The number of people using ChatGPT and artificial intelligence products is soaring. The rate of improvement for basic building blocks underpinning these products is slowing down. The challenge that OpenAI is experiencing with its upcoming flagship model, codenamed Orion, shows what the company is up against. While Orion's performance ended up exceeding that of prior models, the increase in quality was smaller compared to the jump between its last models, GPT3 and GPT4. Some researchers at the company believe Orion isn't reliably better than its predecessor in handling certain tasks. This could be a problem, as Orion may be More expensive for OpenAI to run in its data centers compared to other models. Basically, the idea here is that OpenAI has been training subsequent foundational models with more data, more compute, and it's reached the point of diminishing returns to the point that this grand next model that it's supposed to release, we don't even have a sniff of GPT5 yet. It's calling this Orion is going to be just a bit better, but more expensive. What do you make of this battle?
- [04:32]
Ranjan Roy
I am happy about this. I think I've talked a lot about how I don't need GPT5 just yet. I think the amount of opportunity there is around actually productizing the current models is so massive right now. Again, like, there's so many little magical moments, even with Claude, with ChatGPT, with any of these tools that you see so much potential, but then actually translating that into helping you do your job better or create certain things better. I think that there's. There's just so much work to be done there that everyone competing to kind of create this next massive foundational model has never made a ton of sense to me before. They actually just got a GPT4 or Claude 3.5 opus working, you know, kind of pushing it to its limits and making it work as well as possible. So I'm kind of hoping that this actually moves people towards making tools people use rather than just saying AGI and GPT 5 and whatever, whatever else.
- [05:37]
Alex
Now, look, I hear you, but I also have to disagree. I mean, the field's potential is so much more if these models continue to improve. And while they're good today, they're not where they've been promised to be. And if this is the limit, then it severely diminishes sort of the potential of these models to change everything we do, as the AI industry has been promised. And it's by the way, it's not just this information report. Lots of people have been saying this. So this is Ilya Suscept talked about it. He says to Reuters, results from scaling up pre training, the phase of training and AI model that uses a vast amount of unlabeled data to understand language patterns and structures. It's plateaued. Ben Horowitz and Mark Andreessen talking on their podcast. Horowitz says we're increasing the number of graphics processing units used to train AI, but we're not getting the intelligent improvements at all out of it. And Andreessen has been saying that lots of smart people are working on breaking through the asymptote, figuring out how to get to higher levels of reasoning capability. I mean, shouldn't we just put your concern, your concern about building practical applications aside for a moment? I don't think anyone's going to disagree that it's a time to build practical applications, but isn't this like fairly concerning for the progress of the AI industry if let's say this is about as smart as they're going to get?
- [07:00]
Ranjan Roy
No, I think first of all, I love Ilia Unleashed right now, going on, going to Reuters and now being able to say things like the vast amounts of unlabeled data to understand language and patterns and structures have plateaued. But I, yeah, I think the focus being kind of distracting by focusing only on these step changes in quality of the models has distracted from practical applications. Yes, we should be able to have both, but you even see it in the way that an OpenAI is structured as a company and where they invest their resources. And we talked a lot about this with Corey Weinberg from the information that that the cost structure of OpenAI is still much more heavily towards the R and D and improving the actual models versus building out a good sales force and a sales enablement and customer success team. And these things might sound boring, but if you actually want these technologies to be adapted by corporations and companies and just everyday people, they have to be easier to use and more practical. Like I get the idea that there's a lot of times that if you're using one of these tools, it doesn't work perfectly the first time. And everyone the kind of natural reaction is okay, I guess it's not good enough. But then you learn the better you prompt it, the different you structure your workflow. You can get it to do what you want it to do, but instead I think a lot of these companies are promising. The model will get so smart in the next iteration that you don't even have to do that work around prompting and workflow building that it'll just figure it out and it'll be okay. And AGI will be here, et cetera.
- [08:46]
Alex
But isn't that what they're trying to do? I mean, isn't. Aren't you ignoring the business story here that OpenAI just raised the largest VC round in history, $6 billion, is it? Microsoft or Amazon are hooking up to nuclear power plants. Anthropic is out in the market trying to raise billions of its own. Just from a business standpoint, if these companies cannot advance this anymore, isn't all that money going to come due and sort of crumble the industry?
- [09:11]
Ranjan Roy
I'm not ignoring the business story at all. That is the business story to me, I think like over raising for the R and D side of things rather than the actual like operationalization and building out businesses on top of the existing technology. I mean, again, we've debated this plenty. I think that is a huge mistake in that it actually, you know, potentially hampers the long term development of the industry. So if it actually means this slowdown, you know, puts a little cold water on the promises of GPT5 and whatever else and people just get back to work in terms of actually building things that solve problems, I'm happy about that.
- [09:54]
Alex
I'm trying to pin you down here a little bit on the technology question and you keep wiggling your way out, which I respect. But I have to ask, like, isn't there just a tad of disappointment on your end if this is sort of the end here, the end of the road in terms of where this is?
- [10:09]
Ranjan Roy
Not at all. I mean, the things I've already been able to do. I, I just made a little game in Claude the other day. I saw some video of like, it's kind of like a Space Invaders type of game. I coded a Space Invaders type game with these like custom images myself in an hour and then hit the Claude limit, which a lot of listeners probably do. And it's kind of annoying even as a paying customer. But like, that was magical to me and that exists on the existing technology. It's possible. And there's so many other applications I can imagine if I'm able to do that for fun in an hour that are not being properly explored because all the attention and hype is on the much, much bigger thing. So if the Claude business gets built as, you know, actually teaching people how to use the existing technology well, I think that has again, much better longer term potential than the entire bet on the entire industry is the technology will get step change better in the next year or two.
- [11:17]
Alex
Yeah, I don't know about that. I mean, that has been the bet though.
- [11:20]
Ranjan Roy
It has, it has. And I don't think it's the right one. And I think something that pushes us away from that kind of strategy is going to be a good. It'll shake things up, It'll definitely shake things up. But I think it's, it's healthy for the longer term world of AI.
- [11:38]
Alex
You know, you've really not played into my game today where I wanted to evoke feeling, a one single feeling of disappointment or sadness from you and you say, yeah, it would be tough if this is how I'm feeling. Gosh, like if this promised, you know, AI revolution ends here, then I don't know how far we get. And then I come in and say, well, actually maybe we're not done. You know, sort of like that Walter White give or demo when we say I'm done.
- [12:05]
Ranjan Roy
All right, well, I'll give you. Video generation is the one area that I do think we are severely. We're not even close to anything interesting and we've been promised things that are interesting. That is sora. But we're very, very far away. Even the Runway, MLS and other tools that I've tried were. So that's one area where I see a huge need for technological improvement. But for any content generation, any coding, even data analysis, right now, I think the models are pretty damn good at doing what most people need them to do. We just don't know. Most people just don't know how to use them correctly.
- [12:48]
Alex
Well, Ron Jon, thank you for playing along and I have to inform you, we're not done. We're done when we say I'm done. And that is because. That is because, yes, these research houses might have hit some sort of wall. And the reason why they're hitting the wall is obvious that they're using synthetic data because they've run out of data and it's offering less good results. And this has sort of been the issue with these training, these new models. However, in recent times there has been a development of a new discipline here, which is reasoning. And we talked about it back in the day and that's what sort of freaked Ilya out and he left OpenAI. And that really might be the future, the near future of this field where the models now, such as OpenAI's 01, are prompted and they think. And the more they think, the better they get. And this is again from the information. In OpenAI's case, researchers have developed a type of reasoning model, O1 that takes more time to think about the data that LLM trained on before spitting out an answer. This means the quality of O1's responses can continue to improve when the model is provided with additional computing resources while it's answering user questions, even without making changes to the underlying model. And Casey Newton from Platformer cited one example from one OpenAI researcher talking about it. This OpenAI researcher says, and this was out of TED AI talk, it turns out that having the bot think for just 20 seconds in a hand of poker step by step got the same boosting performance as scaling up the model by a hundred thousand times and training it for 100,000 times longer. So I think what we're about to see is a pivot in the AI research field. Or yes, they might be applying practically some of the models that exist today, but it seems to me like everybody is going to go completely in on this reasoning format and that is going to be where we see the improvements. And that's why I want to highlight this post from Dan Shipper. I saw that I saw this week. He says the message that the information headlines conveys is at odds with what people inside the big labs are actually feeling and saying it is technically correct. But the takeaway for the casual reader that AI progress is slowing is the exact opposite of what I'm hearing. So this might be a combination of spin and reality, but I'm curious how much stake you're putting into reasoning when it comes to being able to advance the status quo.
- [15:21]
Ranjan Roy
Yeah, no, I think both reasoning and how synthetic data is used matter, and I think actually are an almost more promising direction for the industry than just raw processing powder power and size. I think first on the synthetic data, like we're going to be talking about a company writer.com in just a little bit. But they, one of the things they did was like create their own foundation models and they apparently trained them for $700,000 total by using really targeted synthetic data to create different models for different kinds of problems. And I think in the coming months and years we're going to start to see some awkward headlines around size matters because smaller will be better in terms of actual models being used. And like again, one model should not be reliable to solve every problem for everyone at all times versus maybe there is a model focused on financial data analysis and it's actually much better at solving problems around that versus writing poetry or generating images. So I think using really targeted synthetic data for more targeted models is actually a really interesting space in terms of the actual reasoning side. I think that's really interesting. Like, like rather than coming up with new ways of actually generating the answers using the existing information that could be incredibly promising and solve so many of the ideas, so many, so many of the challenges facing the industry like cost for any kind of new foundation model, like just, you know, viability of these things actually succeeding. So I think again today, today I'm positive today these are all good things for me.
- [17:10]
Alex
Right. And the cost really matters because if you're using a reasoning model, a lot of that can happen in inference versus in the training, which is, I think less expensive. Before we move on from this, I just want to talk quickly about this AGI thing that we talk about so often but rarely define and really talk about in context. Right. That all these labs are trying to push toward artificial general intelligence or human level intelligence. And it seems like some inside these, these organizations are like full fledged trying to get there and others, I don't know, probably like see it as useful marketing so they can sell products today. Actually, I think that a lot of the productization that happened has kind of been an accident in places like OpenAI as they've pushed, you know, the research forward. But why don't we just take this point in time to just talk a little bit about AGI? Do you think a it all along has just been this marketing term? And do you think that if we're not going to get there through these current methods that the magic of that marketing falls away a little bit, making it harder to sell into companies, making it harder to fundraise. If all these companies are doing are just sort of productizing what they have today and I guess B. Do you think we'll get there?
- [18:23]
Ranjan Roy
I think I'm going to go with A. And it's because I guess how would you define AGI or artificial? Artificial general intelligence.
- [18:35]
Alex
I think Yann Lecun's definition is really good, which is basically that it's human level intelligence. It can handle a variety of things just the way that a human can.
- [18:45]
Ranjan Roy
What is human level intelligence though? Because there's a. Right. I don't, I think ChatGPT can already do a lot of things better than.
- [18:55]
Alex
But it's, it's almost like yes, it can answer, you know, questions about philosophy the way that a philosophy professor could, but it's almost like the little more nimble things that it really struggles at ChatGPT you can't tell ChatGPT to like, you know, go, you know, write a bunch of emails to people you need to communicate and it does it for you. Well, it's not really able to do that. It's not really able to switch very well between tasks. It's not very well. It's never really able to, you know, learn in the context and get things right the next time. These are all things that I think make human intelligence special is sort of the adaptability and the ability to be, as we say, general. And I don't think AI is there yet.
- [19:34]
Ranjan Roy
Okay, that's, that's a fair definition. And using that definition I actually think it's fine for the industry to not be on the direction of getting there because even what you said, writing a bunch of emails to different people, I think that problem could and should be solved soon in really targeted manners. Like, you know, take your entire existing email history, train something on that, use that to generate new emails and build like a process or workflow where you actually validate them. Like I mean really practically, I think solving that problem could be possible pretty soon and it's just not getting solved because we're still all trying to chase the dreams of AGI. And I think for me, and what exactly it is, the human level reasoning makes some sense. But it's amazing to me that it's always brought up but there's not one like clear accepted definition or one clear vision that's communicated by the biggest people in the field, the Sam Altman's and everything else. So it remains this murky kind of like dystopian robots taking over who knows what it is, will be a line item in a contract with Microsoft's investment in order to change the profit structure. I mean, it's such a nebulous term that that's why I think it does represent a distraction from progress.
- [21:04]
Alex
And I don't know what it says about my life that you're like, imagine the strongest form of artificial intelligence possible. What does it do? And I'm like, yeah, it just writes a bunch of emails like, oh my.
- [21:15]
Ranjan Roy
God, imagine a world where I'm worried about robot takeovers. And you're just trying to go to Inbox zero here.
- [21:24]
Alex
Honestly, if an AI could get me to Inbox zero, it would be a true, a true miracle. I would really believe in the power of science. I would have to get through 12,000 unreads. But it is interesting that they have billed AGI. It initially was like an sort of like what I was describing. Human level intelligence, adept, able to generalize. And now I think it's talked about really in a way that's akin to super intelligence, something that's smarter than humans in almost all fields and can perform things that humans can't. And that's when you hear like the messaging coming out of OpenAI that it can lead scientific discoveries and these type of things. And it's like, okay, that's not really general. That's super intelligence. And, you know, I think that that has led a lot of the investment and a lot of the hype around this that will eventually get there, but it just doesn't seem like it's going to be through the traditional scaling of LLMs. I guess that's my point here.
- [22:14]
Ranjan Roy
Yeah, I agree on that.
- [22:16]
Alex
And you're like, I don't care. That's good.
- [22:18]
Ranjan Roy
I don't care. That's good. That's my new philosophy on scaling laws with LLMs.
- [22:24]
Alex
It's a good tagline. Okay, last thing about this. Did you hear that Google? I think this is worth watching. They have a new experimental Gemini model. It's called Gemini EXP 11 4. And do you know about Chatbot arena, where they test which is the best LLMs? It's currently sitting at the top of Chatbot arena and kicking the butt of ChatGPT4.001 Preview01 mini previous Geminis, all the Claudes. Maybe Google's got it figured out. Whatever they're doing there seems to be working that it's. And this, by the way, folks, this arena is where people compare responses to different models and pick the best one. And it's been voted on by 6,000 folks at this point and is at the very top. So it's quite A moment for Google that I don't want to glance over. And I think we'll probably be coming back to it when we talk about Google's prowess in the field.
- [23:20]
Ranjan Roy
Yeah, I think. And for listeners, I mean, check out Chatbot Arena. It's honestly a fun thing and it's a blind comparison test. So you don't know. You're given two answers, you select which one you think is better and then you find out what's the actual model behind it. So it is essentially an unbiased test. Gemini or. I have a question. Are you using Gemini in day to day life?
- [23:45]
Alex
No, I'm still all in on Claude and I'm looking at Chatbot arena and I'm like, I am not doing it right. I mean, it's interesting because maybe these models can give a better response, but it's also just like it matters, like UI matters. I mean, this is sort of me agreeing with your argument that users matter, personality matters, usefulness matters, even if the model is smarter. But I. This is making me think it's time to give Gemini another shot. How about you?
- [24:12]
Ranjan Roy
I, I don't use it very often. I have it bookmarked, but I mean still. Claude, Perplexity, chatgpt, steady rotation, all for different use cases. I think Perplexity is almost one of the best examples of like UI completely transforming how nice it is to use. And by this point I really thought Gemini should be my entire travel booking, given it's connected to Google travel, Google flights and everything, Google Maps. Like it should be my starting point and it still isn't. And may I still think Google, the answers in the existing form just are not very good. It gets a lot of stuff wrong. It doesn't answer a lot of things as well. And I get trying to be a little bit more conservative and risk averse. I still think Google is incredibly well positioned just given their ecosystem, but it still has not gotten there yet. And I mean it's on maybe the next experimental model, once it becomes reality, will kind of cross that chasm. But we're not there yet.
- [25:25]
Alex
Yes, and I have to say I have become a bit of a perplexity guy. I'll admit it, Perplexity is pretty good.
- [25:31]
Ranjan Roy
It's so good, it's become my kind of companion for other things. Like I think ChatGPT is more when I have like I'm sitting and doing something focused. Claude is for a lot of work, a lot of like more on the coding side and kind of like really the art using artifacts. Perplexity is when I'm watching a movie, a sports game of any sort. Like, like, it just is so good in just giving you quick information in a really nice format with additional links to keep exploring in questions that I think for that like, and which makes me think it's almost and it still is the biggest competitor to Real Search.
- [26:11]
Alex
Yeah, I find it to be really good for research. Imagine trying to sort through a bunch of programs and figuring out what they offer.
- [26:17]
Ranjan Roy
Yeah, I am just making the decision on the Epic pass versus the icon pass for winter skiing. If other if listeners are making the same decision and all done in perplexity and asking like specific questions. Which resorts in Vermont, which resorts in Colorado are there? And like, it was so, so good in doing that.
- [26:39]
Alex
I do hope that perplexity finds a way to get me into China this winter, which I'm hoping to stop in on the way back from Australia. So fingers crossed.
- [26:48]
Ranjan Roy
Where are you looking to go?
- [26:50]
Alex
Beijing. I want to see that wall.
- [26:53]
Ranjan Roy
See that wall?
- [26:54]
Alex
I'll report.
- [26:55]
Ranjan Roy
I was there in 2009. I went to the Great Wall. It was definitely. It was a good time. It was. That's what I'm lived up to, the billing.
- [27:02]
Alex
Nice. So, okay, so let's just take a minute and go through three stories that the two of us kind of found this week that are talking about what you really are interested in, Ranjan, which is the practical application of AI at this stage, and how even if we stop right now, which I don't think we will, we're going to have an extremely powerful technology that's going to disrupt industries and really be practically useful. So why don't you kick it off with this rider fundraising that you talked about? Right, yes.
- [27:30]
Ranjan Roy
So Rytr is a generative AI startup. They just raised 200 million at a nearly $2 billion valuation. What's interesting about them is they built their own foundation models. We had just talked a little earlier about how they build more kind of targeted models that are really focused on solving enterprise business problems. And the entire kind of differentiation that they're focused on is kind of exactly what I've been talking about. Like, they have a lot of big name enterprise customers going in and like, remember, these companies have messy data. They have like lots of really heavy processes that you're not going to just call and make an OpenAI API call and solve. Like, there's so much other work that needs to be done that I think it's. It represents like more the Salesforce Service now world of enterprise software versus OpenAI being. I don't know, just like a more pure heart, a pure tech company, more like. And I think starting to see more companies like that that represent the actual utilization and application layer of generative AI is going to be a good thing. It's going to be a very good.
- [28:47]
Alex
Thing when Writerly does its job. Well, talk about like what you could see it helping a company with Rider.
- [28:55]
Ranjan Roy
Yeah, it's.
- [28:58]
Alex
Look at me, I'm adding the LI at the end of a startup like we're in 2000.
- [29:01]
Ranjan Roy
It's like 2013 again. What a time, what a time. I think what it would look like is going in and actually taking one large enterprise and then recreating hundreds of existing processes and just making them better. Automating some stuff, adding a generative AI layer to other stuff, like maybe keeping some stuff manual, like really rethinking every existing process at a large enterprise and then like actually asking how does generative AI fit into this? And then making that happen and actually creating the kind of frameworks and software that allow you to do that. I think if any company is going to be able to win on that, that's where the, I mean, the value that's going to be accrued there is going to be massive versus just again, I was shocked when I saw OpenAI's kind of like vision around its business is still ChatGPT plus subscriptions. Because I still believe the companies that crack enterprise are going to be the ones that really accrue value in this.
- [30:12]
Alex
That's pretty cool. Okay, so mine is a little bit different. And that is chegg. What AI has done to Chegg. This is just an example again of like how current AI is going to change things no matter what. And for those folks who don't know Chegg is an online education business and they actually started with textbooks and then built up like a pretty serious online education business. And when kids would like, want to research stories or problems, they would, they would say that they were checking it. And this is a Wall street journal story how ChatGPT brought down an online education giant, basically saying that instead of chegging, kids are using ChatGPT now. And Chegg stock is down 99% from early 2021, erasing some 14.5 billion of market value. And there are bond traders. They have doubts that the company will continue bringing in enough cash, pay its debts. So even today we're already seeing this stuff start to really change education. And I think that's like the most obvious place. But we've get, we've already had a few years to see this run its course and look at what it's already done with Chegg. And I think that's like a sign of where things might go with the rest of industry. Not that every other incumbent company is going to lose 99% of their value, but it does kind of show how standard it's become already in education.
- [31:34]
Ranjan Roy
Yeah, I think Also seeing that 99% drawdown brought me back to Chegg was definitely One of those 20, 21, extrapolating into the future, the pandemic. And like anything with the words online education in it just exploded in value. So I think on one hand some of it is related to just that not being the case anymore. But also I think this is actually a really good example. If you think about it like students are the ones, they are. They will take whatever existing technology there is and make it work for them. They will do that work and they will figure out how to answer their homework questions or maybe write a paper in some cases or whatever else it is. And I think like this is a perfect example of, you know, a space where the user is actually driving the innovation themselves because students like free things or cheap things that help them do better more quickly. So that's a good, that's a good use case.
- [32:37]
Alex
User driven innovation. I like that. And it's not the students, it's ad agencies are starting to use it as well. Like talk about, like trying to find the answers to the test. The ad agency, the ad industry is a place where this happens. And this is the last one of the three stories that we're going to talk about in terms of the practical impact today. But again from the Wall Street Journal, AI saves ad agencies a lot of time. Should they still charge by the hour? And this is a story basically that ad agencies have been charging by the hour to clients and now all of a sudden they have ChatGPT. That's made them far more efficient. For instance, you know, if your job, if your job was to write headlines for a brand now instead of having to come up with 50 unique ones, maybe you can write five unique ones and ask to extrapolate out or do the same thing with creative, like creative resizing is now becoming much easier with generative AI. And all those hours that ad agencies spent doing that work, which was really repetitive and not value add, now has become pretty automatable with artificial intelligence. And they're trying to find a new way to charge. And some are going to charge now based off of specific results versus ours. And maybe we see that in places like law. Right. And other disciplines surround. I'm curious, do you think that ad agencies should still charge by the hour given that they were probably not charging for, you know, such valuable work a lot of the time. And now that ChatGPT has all of a sudden made them efficient, they realized that like a lot of the things they were doing weren't really additive to the client. I mean, what do you think the best solution is and what do you think the story tells us?
- [34:19]
Ranjan Roy
I think first it's incredible that you just associated innovation and ad agencies. I think every ad agency out there would be ecstatic that someone else advertising.
- [34:31]
Alex
Listeners shout out to the ad listeners.
- [34:33]
Ranjan Roy
Yeah, I, I think this story is much, much bigger than just ad agencies. And I, I loved this one because I think the pricing of everything that we got used to could change. You just said it. Whether it's ad agencies, law firms will be very similar like outcome based pricing. And in health care, this has been a conversation for years. The idea of outcome based pricing where like the actual results are where you bill rather than the treatment itself is a much, much better way to potentially approach this. So I think for so many of these industries, the way the entire pricing structure changes, it's going to change. And I even think in SaaS that's going to be the case. And there's been a lot of talk around this with even Salesforce's AI agents or in many others is that seat based pricing doesn't make sense in a lot of ways. Like if you're automating a bunch of workflows, how many people are using it is totally irrelevant. So there's definitely going to be some new pricing structure that's something around outcomes, around the amount of compute that is consumed. Like, like it's, it's actually kind of exciting again, like on the productization side of this, like it's going to completely change the way and different industries price and it'll be better, I think.
- [35:55]
Alex
Okay, so after me like sticking up a whole fight about the practical applications of this technology at the beginning of the show, I'm starting to see it your way. I do think that like there's a lot of room ahead in terms of whatever we have today to apply it practically. And I think maybe we should have flipped this like that's actually the big story. And where the models go is sort of now that I'm talking about out loud. I still care more about the, where the models go actually now.
- [36:23]
Ranjan Roy
Are you saying it's time to build?
- [36:26]
Alex
It's time to build Ron, it's time to build. Time to build. Time to break through that asymptote man and just get going.
- [36:33]
Ranjan Roy
Just break the asymptote man.
- [36:35]
Alex
So, on Monday, I spoke with Gustav Soderstrom of Spotify and managed to fit in a question about Parent Mode. But we also spoke a lot about whether generative AI will replace music and whether that is something that can touch someone's heart, whether it was developed by a human or a machine. And, you know, I'm looking through our doc and I see that you have inserted that story back into the conversation, and I'm ready to hear your reaction to what happened on Monday.
- [37:08]
Ranjan Roy
All right, so first of all, and I had asked Alex to ask Spotify CTO CPO about Parent Mode. And my problem is, since I've had kids, my Discover Weekly has been destroyed. I had screenshotted my most recent Discover Weekly when I opened it, and the first song is the Poop Poop Poop song. Yes. And basically everything in there is. Is just something number one hit. It's kind of a banger. But basically, like, not being able to separate out what my kid is listening to versus what I'm listening to makes it. It just destroys the algorithm. And there's no, like, I want to hit a refresh button. He had made an interesting comment that, like, well, making different profiles is actually really bulky and switching. It's kind of a pain. I could make playlists for my kid and then say, do not add these to the algorithm. But I think it's like a reminder that the most complex advanced recommendation system in the world with basic UI problems does not work. And this is another. I think that's another good example of that, that, like, you could have. And I've read stuff over the years, Spotify, how they populate Discover Weekly, and they're very early to machine learning recommendation, but a simple UI problem makes it. So I end up with the Poop Poop Poop song as number one.
- [38:35]
Alex
Yeah. And I do think that they're. They're. It's interesting that they're going to look at these signals and try to, like, get better at figuring out where your listening doesn't match what you usually listen to and try to exclude that. But it seems like a problem that's going to take some time for sure. So enjoy the Poop Poop Poop song. Wheels on the Bus. I was like, we'll. We were talking about it on LinkedIn. I was like, oh, enjoy Wheels on the Bus.
- [38:56]
Ranjan Roy
You're like, no, it's much worse but actually, did you maybe what could solve it? Did you try the new AI generated playlist feature?
- [39:06]
Alex
No, but that's pretty cool. So talk a little bit about that, because that Gustav and I were talking about that, then it came out this week.
- [39:12]
Ranjan Roy
So it's basically you enter a prompt and you get a playlist and you get a bunch of recommended songs and you can kind of like plus plus, plus and choose a bunch of the songs you'd want. And so I literally was like, one of mine was, you are a frat boy in the year 2002 in Atlanta who wants some party songs? And it literally recreated my early call, my college experience, and it was so good. It got the most cheesy, but actually correct and beautiful, beautiful stuff. So then I made a running playlist and I gave it a couple of examples, and again, it nailed it. So it had me start to start thinking, like, imagine if it really can get it to where you just, depending at that moment, are in a particular mood and a really, really, really specific mood, and you just tell the system that, and it creates this playlist for you. And I think this is going to be big for them because I think not everyone is the kind of music listener, which I am, like, who spends time making playlists. So this could really solve this problem for a lot of people.
- [40:22]
Alex
Yeah. And this is what I was trying to speak with Gustav about. It's like, what if you write your prompt in and you actually get AI generated music that will speak to you more than the human generated music? And you actually also dropped this in our document. I was like, what am I looking at here? And it's a bunch of drone video with this, like, really lovely song in the background. And the song you later let on totally AI generated.
- [40:45]
Ranjan Roy
Yeah. I got a drone recently and have been having some fun making some videos with it. And I was up in Hudson Valley in a town called Cold Spring, New York. And literally just with Suno made a prompt, it was like, write a song in a folksy acoustic style about a town named Cold Spring in Hudson Valley and talk about the foliage. So I made this video, put this song as the backing music, shared it with my family in a Apple Photos shared album that we use. And my uncle was like, this is a beautiful singer. Who is she? And then there was that moment of I'm like, do I divulge? And then I did. I was like, yeah, it's AI, which. Which blew some minds. I think the song was genuinely good.
- [41:30]
Alex
Yes, it was really. I enjoyed it very much. All right. And also now that we're talking about Spotify, I'll just note that we are now doing our Wednesday shows via video on Spotify. So if you've recently found the show, this is how it goes. We do Wednesday interviews with folks in the tech industry or outsiders trying to change it. And then on Fridays, Ranjan and I talk through the news. So these shows will be audio only across all platforms. The Wednesday shows video on Spotify. And if you're new here, we appreciate you coming aboard and giving this show a shot. Definitely seen a bunch of new subscribers come in and we appreciate you all. Before we go to the break, just want to say share some gratitude to a couple of our listeners. First of all, Context 1930 shared a comment on Trump in the reviews of the podcast and it was a critical review, but it was five stars. We're taking it into account and we appreciate the way that you shared that feedback. It helps us and it helps the podcast and I think that's the best way to do it. So thank you, Context 1930. Also, Luke Squire made a comment about our discussion about polling versus prediction markets on LinkedIn. Basically in favor of polling versus the prediction markets. We've got a couple of those. And that's another great way to share feedback and thoughts on the show is shared on LinkedIn. And critical or not, we love to hear what you think about the show and it obviously gets the word out to others. So we appreciate that. Thank you, Luke. And then Graham High emailed me with our on our email address for feedback, which you can find in the show notes and made a very interesting point. So we talked a couple weeks about how the government should build its own starlink. We talked about that a few weeks ago and Graham pointed out, and I'm embarrassed to admit that I didn't know this, that the Department of Defense has actually already started work on its own satellite Internet company or communication system working with SpaceX. It's called Starshield. Ranjan, did you know about it? Here's from one story about it. It's a militarized version of SpaceX's Starlink Internet satellites with enhanced encryption and other security features. And unlike Starlink, which is a commercial service, the starshield satellites would be owned and controlled by the US Government. So the government is actually building this?
- [43:50]
Ranjan Roy
I did not know that, but I think we need to do more space coverage.
- [43:55]
Alex
That's right.
- [43:55]
Ranjan Roy
I think. I think space for 2025 is going to be a good topic.
- [44:00]
Alex
All right. Bezos put us in a spaceship. We'll take our mics and we'll do it.
- [44:05]
Ranjan Roy
And podcasting live from Blue origin. That's right, 2025 goals.
- [44:10]
Alex
Yeah, Bezos, we know you listen, so just do it. All right, let's take a break. We're going to talk about Blue sky and if we have time, we're going to talk about Apple Smart Glasses right after this.
- [44:21]
Tomer Cohen
I'm Tomer Cohen, LinkedIn's chief product officer. If you're just as curious as I am about the way things are built, the insight behind what it takes to create world renowned products, then join me for my new podcast, Building One. Together we'll get to learn from leaders around the world, people with diverse backgrounds across multiple industries. Each will share insights into their craft. So listen and follow my new show, Building One on Apple Podcast or wherever you get your podcasts and check out the conversation on LinkedIn. It's going to be great.
- [44:54]
Alex
And we're back here on big Technology Podcast Friday Edition. Just a few minutes left, but I definitely want to talk quickly about this Blue sky surge. So Blue sky is now up to 15 million users and it is, it's really soaring in the wake of the election. I don't know about you, but I've definitely noticed myself and lots of other folks have talked about how they've seen mass amounts of followers delete their Twitter accounts. And I think Blue sky and Threads, which threads has added 15 million users and just since the start of November, have definitely benefited from this. So do you think that this has staying power or is it a flash in the pan?
- [45:31]
Ranjan Roy
I think it does have staying power this time. So I went back to the Blue sky account I'd created like a year and a half ago maybe. And it was interesting. I actually saw people who I would engage with on Twitter all the time who I hadn't really processed, had left, but just kind of hadn't thought about or noticed in a while. And suddenly I was like, oh wait, they're, they're alive and kicking and just having those same conversations, especially around a lot of like economics topics, finance topics, even in tech as well. I found a lot of the, a lot of tweeters from my past in there. So I think, because again, from a product standpoint, from before, even like creating an account signing in following was kind of a pain. And then now when I went back, it's, it's pretty much on par with Twitter, slash X. And so I think there's, there's staying power here because again, the actual technology behind any of these apps is not that complicated. It's purely about the content and the people involved. So I think it does represent a risk this time. But we've said this a few times now, so.
- [46:43]
Alex
Yeah, and I'm about to pour some cold water on this. Max Reed, who writes Read Max on Substack. He says, from what I can tell, the users who've been joining Blue sky and Mass recently are members of the big blob of liberal to left wing journalists, academics and lawyers and tech workers, politically engaged email job types who were the early Twitter adopters and whose compulsive use of the site over the years was an important force in shaping its culture and norms. But he says Blue sky is really acting more like a large discord server, a place to socialize, bullshit banter and kill time than a proper Twitter replacement. So basically what he's saying is it's inhabited only by those people and it feels a lot like the old Twitter, but it just doesn't have the user numbers that it used to have and therefore the Blue sky boom might be an illusion. What do you think about that?
- [47:35]
Ranjan Roy
No. So I think when I had gone on it way back, it was like the extreme version of the anti elon Musk anti Twitter types. This time there's a lot of sports highlights on there which could be my, I mean there, there, there's more kind of normie content on it this time around and I pretty quickly was able to find a lot of good follows. So I think that it, that's still kind of how things were and maybe that's his specific feed, but I think it's different this time.
- [48:08]
Alex
Let's see. I don't think that it's going to work. One thing we can tell say for sure is it doesn't look like Threads is working. I mean threads added 15 million people since the start of the month and Michael Earmonth, who I work with as an editor, pointed out to me, he's like, does it feel like that? No, it feels like the same thing. It's just people complaining about threads.
- [48:31]
Ranjan Roy
I can't with threads. I opened it up again and yeah, I mean it's so odd in terms of like. And I've tried to follow a bunch of people on it, but I don't know, it just does not deliver more real time. Interesting conversation, I will admit. Blue Sky, I actually moved it to my homepage on my iPhone and moved X off of it. And then in New York this week, on Thursday, we were looking out our window and you saw smoke coming out of a building in Midtown, I don't know. Did you even hear about.
- [49:11]
Alex
Yes, of course.
- [49:12]
Ranjan Roy
So there was a fire in Hudson Yards apparently it was like a mechanical room blew up or something like that. But. And no one was injured or anything, but. But I actually tested. I went to Blue sky and searched NYC fire. Nothing. I went to Threads. Nothing. I went to Twitter and got all the info I needed right away.
- [49:33]
Alex
Yeah, that's why I think Twitter is going to be the one with staying powerful. Just the network effects. It's very, very. It might be the most difficult to replace social network like we've seen like blue Facebook, at least in the US start to lose a lot of interest. People are on Instagram now. I just don't see it happening with Twitter because it is just the group of sickos that have been on that platform and the network effects there is very difficult to displace. Okay, last story of the day Apple is thinking about smart glasses was in our doc and for like the last week, but there's been a lot of politics to talk about. This is from Bloomberg. Apple is exploring a push into smart glasses with an internal study of products currently on the market. The initiative, code codenamed Atlas, got underway last week and involves gathering feedback from Apple employees on smart glasses. And it's been led by Apple's Product Systems Quality Team, part of the hardware engineering division. So it's very interesting to me that smart glasses are already becoming a thing. Meta has a great pair out with the Ray Bans and Apple has been beaten to the punch here. And I think it's not going to be too long until we see Apple build a product like this of their own, if not one with an enhanced Siri to hit one of your most favorite things. What do you think?
- [50:54]
Ranjan Roy
Yeah, I think, I mean, I told you a couple of weeks ago that I'm testing the Snapchat, the Snap Spectacles, which is their new augmented reality glasses that you can get as a developer versus Orion from Facebook. The actual AR glasses are not available for any kind of like general release. But after using the smart, the spectacles, it's, they're amazing. And I talked about it like even my son can use them instantly. My mother, like anyone of all ages and kind of like technological proclivity can just pick them up and use them. And I think this is the, this is the form factor of the future. This is like what we're, we are going to all own some kind of glasses and Apple's got to get on there quickly. And the Vision Pro was not that. And VR.
- [51:48]
Alex
Yeah. What do you think it says about Apple that they haven't been able to do this? It's not a good sign.
- [51:54]
Ranjan Roy
No, it's not a good sign. I think like Apple Intelligence. I mean, if you think about it, we have a bunch of misses in a row. Apple Intelligence, maybe it'll come around, but it is so far from anything we have seen even remotely close to useful. The Vision Pro flop. I mean, and I'm still upgrading my Macs and AirPods and iPhones and all that, but it's just, it's. I mean, and again, at their scale, they need to find that next big winner. We, everyone knows it and it does not feel. Maybe Siri will work in a few.
- [52:34]
Alex
I think Apple's best chance is that Mark Zuckerberg gets so ahead of himself on his rebranding campaign where now he's like, you know, cool MMA Zuck with the chain and the big T shirt and the long hair that he distracts from the mission and then gives Apple an opening. And you know, I'm a fan of a lot of Zuckerberg's side projects, but there was one this week that I just didn't think hit and raised some emoji red flags for me. And that was a collaboration that he did with T Pain to sing a song, Low. It's called Low. It's one of the hit songs back in the day.
- [53:19]
Ranjan Roy
It's Get Low, Get Low. Alex. It. It might have made it into my Spotify AI playlist from 2000s college party music.
- [53:29]
Alex
And he worked with T Pain to record a version of this song. It's quite X rated and T Pain wasn't even involved in Get Low back in the day. But this is from Business Insider. The duo, which calls itself Z Pain, released the Slow Down, Not Safe for Work track on Spotify and it features a heavily auto tune. Zuckerberg syncing original lyrics about going to a club and getting confronted by a security guard. And it features Zuckerberg singing some lyrics that I really never wanted to hear him sing.
- [54:03]
Ranjan Roy
Oh, this was awful. I mean, what I kind of love is thinking about like trillions of dollars of market capitalization potentially swinging on Mark Zuckerberg sitting down with T Pain with an acoustic guitar. I can't remember, does he actually play it in the video or. But singing.
- [54:26]
Alex
I'm trauma. Wiping it from my head. Yeah.
- [54:28]
Ranjan Roy
And singing about sweat in the nether regions and with. And to me, the most ridiculous part is that as you said in the Business Insider article, said T Pain didn't even Sing get low. It was Lil Jon and the east side Boys back in the day. So, like, just how this came to be and what this could mean. Like, you're. You're going around laying off people, telling them this is the year of efficiency, and then you're trying to call yourself Z Pain and come up with some weird alter ego and sw. Oh, man, I. I can't with this one. This one was too much.
- [55:10]
Alex
Ranjan. I think there's only one thing that's left to do at this point.
- [55:14]
Ranjan Roy
What's that?
- [55:15]
Alex
That is to queue up the song and play about as much of it as we can get away with without being kicked off of the podcast platforms. So thank you for coming on the show, Ranjan. Thank you, everybody, for listening. And now to play us out, Z Pain, Mark Zuckerberg, and T. Pain. We'll see you next time on big technology podcast three six nine.
- [55:37]
Michael Kovnat
Damn, you're fine Hoping you can sock.
- [55:41]
Ranjan Roy
It to me, baby one more time get low, get low get low, get low, get low get low get low get low, get low, get low from the windows to the walls Till sweat drops down my balls Till all these.
- [56:05]
Michael Kovnat
Crawl oh SK oh skeepsk God damn.
- [56:12]
Ranjan Roy
Oh SK oh SK Godamn.