← Software Engineering Daily
Software Engineering Daily
SED News: Bezos Returns to Building, AI’s Reality Check, and Europe’s Cloud Ambitions
00:54:01
SED News is a monthly podcast from Software Engineering Daily where hosts Gregor Vand and Sean Falconer unpack the biggest stories shaping software engineering, Silicon Valley, and the broader tech industry. In this episode,
Loading summary
Transcript75 lines
- [00:12]
A
Hello and welcome to SED News. I'm Gregor Vand.
- [00:16]
B
And I'm Shawn Faulkner.
- [00:18]
A
And this is, as I'm sure many of you know by now, but just slightly different format of SE Daily where we cover tech headlines. We have a sort of main topic in the middle. Today's topic is tipping points of technologies. And then we are going to roll into Hacker News Highlights, where we just take a few of the fun, weird and interesting things that have popped up on Hacker News that you may have missed, which is very easy these days given how much lands on there. But as always, I mean, we're heading to the holidays. We're recording this the day before Thanksgiving. Certainly I would love to celebrate Thanksgiving. It's a great holiday. Unfortunately, I tend to live in countries that doesn't really get celebrated, but celebrating it. Right, Sean?
- [00:59]
B
Yes. And as a Canadian that's now lived in the US for 15 years, I've migrated from the Canadian Thanksgiving, which is celebrated in October, to the U.S. thanksgiving, that's celebrated in November. So I've done the full migration, essentially.
- [01:14]
A
Awesome. And yeah. Anything else that you've been up to, Sean, in the last few weeks since we last recorded?
- [01:20]
B
Oh, nothing specific. I mean, the big thing that's coming up for me is the week after Thanksgiving is AWS Re Invent. So I'll be going, heading off to Vegas very early Monday morning to be there for the week, which is going to be a pretty Busy Week with 60, 70,000 various people in the technology industry. So should be fun. How about you?
- [01:39]
A
Yeah, I've got a few colleagues heading to that. Yeah, just actually, for those that are listening, I'm sure there's a lot of people that have never been to a Re Invent and maybe will never get the chance. Can you describe what is a Re Invent to someone that doesn't even.
- [01:53]
B
Yeah, I mean, so this is aws, Amazon's big marquee technology event. It's pretty massive. It's not the biggest tech conference I've ever been to. I think Mobile World Congress is, which might be the biggest tech conference in the world. Whether that's like 150,000 people, but I know 60,000 or so. Registration 70,000. So it takes place in Las Vegas across multiple hotels. It's always the first week of December, probably because AWS got some deal there a decade ago, even though it's kind of inconvenient for people. But there's just a ton of stuff going on. They make a lot of big announcements there. There's so many parallel Tracks of talks. I'm giving a talk there. But on top of the registrations, there's Also probably another 10, 20,000 people who are just there, who don't even register just for the side events because there's just a continual thing going on. And so many people that you would want to converse with in technology or meet up, do business with are all there, all kind of centralized around this. So there's a lot of just networking and community stuff going on all the time.
- [03:00]
A
Yeah, exactly. This wasn't re invent, but a startup. I won't name the startup, but we did actually cover them on SE Daily. But they said that they went to a conference in Vegas recently and rather than pay the 20 grand for a booth as a marketing thing, instead they said, hey, we're going to have 20,000 in credit on blackjack table and you can come play your hand so long as you take a meeting with us. Basically. Which I thought was a pretty cool way to sort of leverage Vegas as a meeting point.
- [03:32]
B
Yeah, I mean the Expo hall is. I mean, it's like you're in an airline hangar or something like that. It's just so huge. It's pretty hard to stand out. I always felt like, because the booths are expensive. Unless you have the means to pay for a gigantic booth where you can really stand out by paying for it. I think if you want a booth, you go with a small one because there's not that much, I think, to gain from having sort of the slightly larger booth versus the small booth and the amount of money that you're spending on it. You could probably spend it on, I don't know, giving people credits at a blackjack booth or having some coffee or drinks or something like that.
- [04:10]
A
In your booth. Exactly, yeah. On my side, just heads down work, unfortunately. But I'm looking forward to the Christmas holiday season, which is coming up soonish. I'm going to head back to the UK next week, see some family. Shorter visit than usual, but it'll be nice to get some cold weather. Right. Moving into the headlines. These are the main tech headlines. They might hit your normal mainstream news. Things like Wall street journal, Financial Times, TechCrunch, this kind of thing. The first one we've not heard about Bezos in a while, kind of. Apart from things like weddings and frivolous things like that.
- [04:46]
B
Yeah, mostly personal things.
- [04:48]
A
Exactly. Yeah. But what is Bezos actually doing? Jeff Bezos of Amazon, actually, from a sort of work standpoint, he's not exactly unemployed, but. Well, he's just announced that he is stepping back into the CEO role of something called Project Prometheus. And they say that this is a company focusing on artificial intelligence for the engineering and manufacturing of computers, automobiles and spacecraft, which is pretty broad brush these days, but kind of, I guess there's something clear there. Where it's AI but kind of for deep hardware is kind of what I took from that.
- [05:24]
B
I did a little bit of research into this. There's not a ton about the company yet, but it sounds like it kind of relates to something that we had talked about previously, which is AI applied to physical tasks. The intersection between AI and learning from the physical world. Large language models were trained essentially on digitized textual information. And there's a bunch of companies now that are looking at how do we train models based on physical interactions, simulations with the physical environment for things like either robotics, sometimes scientific discovery, drug design, all kinds of different things. So it'll be interesting to see which direction they go. They have $6.2 billion in funding, so.
- [06:09]
A
They have some Runway that will cover Jeff's salary hopefully.
- [06:14]
B
Exactly. And his co founder was the founder of Verily, which was one of the. It was like a healthcare spinoff of some of Google's Moonshot projects.
- [06:25]
A
Ah, interesting. Okay. But yeah, I mean I think this is. Although he has Blue Origin, the space exploration company, he was never actually the CEO of that company. He was obviously a founder. But he did say at the time and has said again that Amazon was his focus and he's always sort of trying to driver wedge between him and Elon Musk I think, and sort of pointing out that he was actually being the CEO of Amazon and letting other people run his other companies. Whereas Elon Musk of course tries to be the CEO of many companies at once and some might say loses focus. So this is the first time that Bezos has stepped back into CEO after stepping down from that role at Amazon. So yeah, could be interesting. I think a quote from Wall Street Journal was just that it looks like Bezos is back to building and has a new list of ideas on his whiteboard. He's also known for being the prolific idea creation person. Too many basically is usually what the problem has been with Bezos. Too many ideas and his teams don't know what to do with all these ideas. So let's see if Prometheus can work with that.
- [07:31]
B
Yeah, I mean I think we're seeing a lot of well known people in the tech world sort of be drawn out of semi retirement or whatever they're doing by what's happening In AI right now. Sergey Brin did his first pull request in however many years this year. So there's a lot going on in the space. It makes sense. It's an exciting space. So people who've made their careers perhaps in other ways, there's probably enough of a draw to bring them back out of semi retirement to do something interesting.
- [08:04]
A
Yeah. So moving on, there's been a lot of headlines. So not exactly targeting one company specifically, but just a general AI. Is it actually going anywhere? Is it costing too much money? So there was an article in the Wall Street Journal again, AI investors want more making it and less faking it. So just this idea that companies are continuing to, in net terms, lose money on their AI gambits, but ultimately are sort of making it up on volume of could be revenue or users and so on and so forth. But yeah, what did you make of this one, Sean?
- [08:44]
B
Yeah, I mean there's always people say losing money, but making up a volume kind of joke in the Valley and stuff. But overall, I don't know that I really agree with a lot of the points in the article. I think that assumes things like inference costs being static ignores that likely compute costs will go down. Just like the cost of transistors have gone down, compute has gone down over many decades now. So as compute goes down, of course with scale, then there's more opportunity to make money. One of the things the author talks about is he argues that investors may conclude that they don't want to pay these big costs to get to some uncertain point of whether AI is going to work or not. But I would say for the most part, the reality is a lot of these companies have no choice. If Google stopped spending tomorrow on AI infrastructure, but Microsoft continued, then Google risks essentially ceasing to exist at some point. Because I think spending right now in part is driven by this existential fear of becoming irrelevant. No one wants to become, I don't know, the Xerox of the emerging AI world. And they talk about this concept of faking it until you make it. But I would say AI isn't faking it, at least not in totality. There's certainly a certain amount of hype and things that aren't necessarily real. But AI is today at a point where it's writing quite a large percentage of global code. It's handling a lot of tier one customer support, generating media, and even outside of, we go outside of generative AI and search for large language models, machine learning. And AI has been used for very, very long time successfully in Software, it's just generative. AI has really brought it to the forefront. But behind the scenes, black box ML has been used in many, many industries for a very long time, successfully. So I think it's a little bit hyperbolic to kind of say that it's all fakery essentially.
- [10:47]
A
Yeah, I agree with you. Generally. I think the part that is tricky to sort of merge in with that is there are unfortunately quite a few startups that do look quite snake oily at the moment. I mean this always happens. You have a, you have this sort of boom technology, everyone jumps on it and just people still tagging just AI onto the end of a company name. Instead of the dot com boom bust, if you want to call it that. Everyone just said I'm dada.com, and now everyone's like, I'm so and so AI.
- [11:20]
B
I think it's consistent with any early thing that becomes hot. There was a period in the early days of the gig economy when Uber and Lyft and companies like that were taking off and then suddenly everything was like Uber for nurses.
- [11:33]
A
Exactly.
- [11:33]
B
Or if you remember the SOMO local or whatever social mobile local era as well, which is based on companies like.
- [11:44]
A
Yelp, like Foursquare and that kind of thing.
- [11:48]
B
Foursquare being successful. Yeah. And then everything had to have a social mobile local component. When social like Facebook and other social platforms blew up, everything suddenly had to have some virality social component, whether made sense or not. So to me, if you've been in tech long enough, you've seen sort of these cycles and this feels similar to me.
- [12:12]
A
Yeah, and we'll kind of get onto that in the main topic where we look at tipping points, that is when and why do sort of technologies end up getting the adoption and the respect you could even say that they deserve. So we will get onto that. Yeah. Following on from last month's main topic, we were looking at kind of the interconnected web of the AI players. OpenAI was the center of the web, if you want to call it that. And yeah, you turned up a pretty interesting article Sean called Is OpenAI screwed? So just straight to the point. Yeah, although this wasn't sort of main mainstream news, but this was a very well tons of comments and it was on Medium, like a lot of likes. And you can almost say Medium is basically a sort of semi news outlet at this point. So what was kind of going on in this article?
- [13:03]
B
Yeah, I think it's a very well researched, well written article that dives into this idea of investigating what's happening with OpenAI. What are the existential threats to OpenAI and some of the moves that they're doing? They protect themselves against that. So the author argues that they're doing a lot to make themselves too big to fail by intertwining them with a lot of other companies. And we talked a little bit about this the last time as well. For a long time, I thought in a similar way, where OpenAI could be in trouble in the long run versus hyperscalers, because companies like Google have this very large existing business to fuel their AI investments. They have data centers, they have TPUs, they have the talent, they have billions of users across thousands of different products that they can immediately, immediately bring AI to. And that felt like an unfair advantage in comparison to being a dedicated model company. But OpenAI has done a lot, at least from an outsider's perspective, in the last year to try to, I think, protect themselves essentially against that, to be too big to fail. I think some of the arguments in the article don't necessarily hold up. They try to make the case that OpenAI is burning so much money, $12 billion quarterly loss, that is proof that the business model is broken. But if that $10 billion of that 12 billion was spent on Nvidia H2 hundreds or something like that, that's not really lost money. You're essentially converting that money into compute, where that compute is tremendously valuable. You could even argue that that's part of their competitive moat, is to own all that compute and then turn it into money later. So yes, it is spending, but it's not necessarily like, hey, we just lost this money, it's irrecoverable.
- [14:52]
A
I think that's a good read on that. I think it's just the volumes, right? It's just the volumes. No one is disputing that spending a lot of money to make it up later is sort of, that is tech. Often it just seems like OpenAI, as you're kind of saying, is almost making themselves too big to fail. The numbers just keep growing exponentially without actual income dollars to back up the supposed spending. And again, when you say spending, we're kind of talking about deals on paper, committing to things. Where is the money going to come? That's kind of like debt in a way. Call me old fashioned, but that is kind of debt where you're just saying like, hey, we promise to pay you if you give us this stuff and we'll figure out how to pay you the actual money later kind of thing. Tired of babysitting autoscalers and Overspending on cloud costs. Meet Thoris, the platform that makes engineers heroes to their finance and business leaders. Thoris intelligently manages kubernetes clusters automatically, right, sizing and scaling workloads while preventing downtime from traffic spikes. It anticipates usage and capacity needs so systems stay fast, reliable and efficient. Without constant tuning, teams using Thoris cut cloud spend by 40 to 60%. Thoris predicts compute and GPU demand before it happens, keeping performance smooth and costs in check. Stop wasting, compute and guessing your resource needs. Let Thoris handle your auto scaling so your teams can focus on building. Find out how much you can save with Thoris. Visit Thoris AI and try our cloud savings calculator.
- [16:27]
B
I think some of that, although the numbers are really big and maybe that's why it's a little bit jarring, but if you look at Amazon, lost money for a very long time, was in the red for essentially eventually they had to get to the economies of scale to a crazy point in order to start making money. And then Facebook had no monetization in it for years before it was all about gathering eyeballs. So they were eating all that infrastructure costs, costs, engineering costs to scale that platform before they turned it into this money making machine. But it took quite a while to get there. I think you can make the same argument that OpenAI is kind of doing on a similar trajectory. It's just the numbers are so astronomical, but it's hard to. You're talking like tens of billions of dollars.
- [17:11]
A
And the article also touches just I guess on Sam Altman himself.
- [17:16]
B
Yeah, and obviously I don't know the guy at all, but I think if you look back at history, we were talking about Bezos, Bill Gates, Steve Jobs, all these people, built these gigantic companies, these kind of mega empires, become some of the richest people in the world. They're all notoriously difficult to deal with. And I think that it probably takes a certain maniacally driven individual to build these monster companies. And with that probably comes a certain cost of that person sometimes being difficult to deal with.
- [17:49]
A
Yeah, absolutely. Again, it's going to be interesting. Every week something is popping up at the moment which is sort of touching on what is actually going on over there. OpenAI. So yeah, I guess it will remain to be seen. Let's see where we land on this in a month's time, because a month is sort of like six months of what used to pre AI time. So yeah, so let's see where we land on that one. Finally, just kind of wrapping up on the headlines. Yeah, this was just TechCrunch touched on the fact that it sort of looks like boom time in the Nordics. I would say again for tech Nordics being Finland, Norway, Sweden, Denmark and yeah, if I think pre AI, the Nordics for me are very much like Spotify, Klarna, Skype even, which I think is actually Estonian. They do kind of class themselves in the Nordics as well. So it's been this kind of interesting place outside of the Silicon Valley SF ecosystem that they've often produced some really, really strong startups. Not nearly as many, but they have produced some of the strong like Spotify. Everyone knows Spotify. I can't say there's a single Spotify that came out of the uk for example. I've always chalked it up to just it's a very dark place half the year and got a lot of very smart educated people and they've got nothing else to do other than sit and program things. But this article also touches on the idea that well, today we've got Lovable is one of the fastest growing startups of all time. That's from Sweden and it's definitely the epicenter of this. Sweden and Stockholm have become now this kind of new second hub of AI, I would say, and AI startups.
- [19:41]
B
Yeah, it's interesting. I mean I think the Nordics hit rate on companies is pretty high given their population relatives to say, like the US. I mean I think there's less than 30 million people that live across all the Nordic countries, but Finland in particular has a very strong education system. And I also think that once a certain industry becomes successful in a country or an area of the world, then the art of the possible has been demonstrated and people follow that. It's kind of like if you're going to be an actor, Los Angeles is the place to migrate to and everybody's in that industry. So you create these bubbles of industry. Of course in the US it's predominantly the Bay Area for technology and startups and things like that, but you're so surrounded by it, then it becomes something that as a kid you see like okay, well I could do that. Why not? So maybe they have something like that going on in those countries.
- [20:36]
A
Yeah, and I've actually spent quite a bit of time in Norway, so I can maybe sort of relate to this. It touches on the article that the regions social safety that lets young people take risks without fear of losing everything. I do kind of agree with that. I would also take the other side of the coin, which is because it's so Safe and salaries are pretty good in these countries generally, I would say it's still 50, 50, whether that incentivizes you to go off and start your own company. Okay, strong safety net, equally, you can get paid quite well, quite frankly on many, many jobs that in many other countries you just wouldn't maybe consider because the salaries just don't make sense. We're talking things like teachers and even working in a supermarket in Norway is not like you can make a completely okay life that way. And that's a very nice way to run a country. So yeah, I think it's sort of interesting looking at that side of things. But I do agree to some respect. I would say that more so than say Singapore where people are quite afraid to fail. And I think that does drive the sort of how startups happen here. They tend to be often people that are, to be frank, coming from quite privileged backgrounds because they can afford to try it, whereas a lot of people simply cannot afford to even try in places like this.
- [21:53]
B
Yeah, I think there's a lot of things that go into that because I don't think it's about necessarily do you have a safety net or is this your path to making money? Because if you're going into doing a startup for the money, you're going to be most of the time pretty disappointed.
- [22:09]
A
Please don't do that.
- [22:11]
B
There's a much easier path to financial security than trying to do your own startup. So I think there's more of either a culture of innovation and also like you had spoken to of are you okay with the prospect of failing? Because there is risk that you take on in terms of like, hey, I'm going to go do this thing. I might not make any money from it. It might fail massively and I might s multiple years of my life to try to make that happen and it might not happen at all and be sort of okay with that.
- [22:40]
A
Yeah. And just a sort of sidebar on the Nordic side of things was yeah, Lovable just announced acquiring Malnet. So this is a very young company, about 2 years old. But then so is Lovable I think or Lovable is even less. But yeah, Malnet just three founders. It was aiming to be Europe's cloud provider because I think there's been a strong leaning away from having to rely on US infrastructure and around the data privacy that sort of comes with that, especially if you don't live or none of your users are based in the US however they were moving into hosted postgres. I think we've seen this movie before and yeah, so Lovable's just acquired them and announcement was simply that the team and technology is being rolled into Lovable and the team will continue to work with them on that one. So that's kind of interesting. Just another potential sort of hosted postgres roll up there. So moving on to the main topic, we're kind of calling it Tech tipping points. So basically how long does a far out technology sort of take to be adopted? And there's many technologies we could touch on here. We're framing this obviously around AI. Why are we doing that? Because not just to talk about AI yet again, but people are asking is there a bubble in AI? People come and ask me that. And a bubble is a more financial question. We tend to look at it more from the technology side. So we're less interested about in this case. We're less interested in the deals being done and valuations and all that. We're more interested in. Sure. But if there is actually a technology behind it that is worth pursuing, at what point do people kind of stop saying oh it's nonsense and this will never work? And at what point did we kind of see these technologies actually get respected and adopted? So we'll look at things like the Internet is the obvious one that came. Internet obviously predates AI and how did that look? We'll also touch on things just like smartphones in general. We look at alternative services like Uber. We've already touched on it slightly briefly in this episode. There's also kind of these moonshot technologies that are starting to look less moonshot y now. Like autonomous driving. So I guess if we sort of look at the Internet to begin with. The Internet through the sort of bubble years, people just couldn't really see what it was going to deliver for them for a long time. And even E commerce took a long, long time to be taken seriously quite frankly. But I think this is probably running theme here is there have to be so many pieces of the puzzle that come together to make these technologies. Sort of makes sense. I'll take a slightly more nuanced example just to begin with which is actually Internet on planes. So if we take not just the whole Internet but putting Internet on a plane. This has been around for like 10 years now. Ish. And airlines were starting to kind of roll it out about then with these Panasonic boxes. And then it was like viasat which was a slightly different way of looking at ground versus satellites. But crucially the speeds were terrible. Connections, very patchy when you're flying. There's a Map showing you, oh, you're going to fly over this country now and it will drop out and then it will come back again, which is just if you're actually expecting to work on a plane that doesn't work.
- [26:05]
B
Yeah, it's worse than the Internet in the 90s for the most part.
- [26:09]
A
Right, Yeah.
- [26:10]
B
I regret buying Internet access on planes most of the time.
- [26:15]
A
Exactly. I actually only try Internet on a plane if it's free and again I was on, I think it was Japan Airlines flight and they said, oh we've just installed WI fi and please try it. And I did and it didn't work. But what's happened is Starlink, Starlink's come along and obviously Starlink is disrupting a lot of technologies and even countries quite frankly, because Starlink has suddenly breaking down barriers when it comes to censorship. And as we just touched on in another episode with one of the other hosts, drone warfare has even changed because of Starlink. But in this case, Starlink on a plane giving actually good Internet speeds, actual reliability. So it's less about is Internet on a plane useful? It was about is Internet on a plane reliable, fast? Is the Internet that we expect on the ground, in the air? Qatar Airways have rolled out across a whole fleet of their planes and the consensus is it's amazing. I'm going to be flying on Qatar next week. I hope at least one of the planes has Starlink. So I'll report back and see if it has. Actually I'm genuinely planning to work on that plane and for me that's the tipping point. Like the tipping point of can I actually say, hey, I'm going to work today on a plane as opposed to, you know, these are like seven, eight hour flights like, like can I work on a plane versus just saying, nah, I've got to take the day off because I can't guarantee I can do any meaningful work.
- [27:40]
B
Yeah, Capital One's tech team isn't just talking about multi agentic AI. They already deployed one. It's called chat concierge and is simplifying car shopping using self reflection and layered reasoning with live API checks. It doesn't just help buyers find a car they love, it helps schedule a test drive, get pre approved for financing and estimate trade in value. Advanced, intuitive and deployed. That's how they stack. That's technology at Capital One. I think Internet, that's a good example of something where everyone can agree that if you can make the technology work, there's value there. So the number one risk is a technical risk. It's the kind of things that the VC firm Engineering Capital invests in, their number one thesis is they invest in companies that focus on solving some sort of engineering challenge. And if they could solve that engineering challenge, then it kind of sells itself, essentially, because the key unlock is like, can you get the technology right? I think WI fi and airplanes are a really good example of that.
- [28:41]
A
And yeah, I mean, airlines could not have predicted Starlink. Quite frankly, Starlink is very much linked to Elon Musk. I'm sure there's a ton of other big minds there that have made this possible, but it's the kind of thing where you've had all these companies like viasat working on this stuff for years and years and years. And then Starlink just kind of comes out of almost nowhere and wasn't even designed specifically for this purpose. It was just, hey, if this is possible, this is going to be a game changer. It was possible with a lot of investment. And here we are. So we'll move on to AI and Internet together. How do they compare? Contrast. But if we just look at a couple of others before that, I guess we could almost take these two together. I would say, like smartphones and things like, when I say alternative services, there were taxis and then there was Uber and then there was all the sort of geographical derivatives of Uber. But this was it. Again, people couldn't foresee how taxis could get disrupted. But it did rely on there being a device in your pocket that could do all that stuff, basically.
- [29:48]
B
Yeah. I mean, the fact that you had a computer in everyone's pocket allowed you to get rid of the proxy service of the taxi dispatcher and you just go point to point and then essentially the dispatcher is the software itself. And clearly taxis have been around for a long time, didn't see smartphones coming and being that disruptor. But I think that's probably a little bit more similar in some ways to the Internet on an airplane than I would say, the creation of the Internet or what's happening in AI. Although here you're sort of disrupting an existing industry. So it's changing the pattern of how somebody uses a service. So it's a little bit different. But again, it's like some technology unlock that led to that disruption.
- [30:35]
A
Yeah. And as we get onto, I think the taxi dispatcher, that's interesting because that's like a job role that effectively has mostly disappeared. I mean, of course, I would say probably in London, New York, there are still taxi dispatchers. And it's not like that was a massive job Generally, but there were tons and tons of dispatchers. But it is a job that has effectively been made moot. And as we'll get onto with AI, that's the big fear around, people keep saying AI is nonsense and it's going to take all our jobs.
- [31:10]
B
I think too, there was companies that tried to do the version of Uber, but work with the existing taxis, but they could never really make them work in a reliable way because of essentially having this person in the middle of the communication. So you just didn't get really up to date information on whether the car was coming and it just ended up being extremely unreliable. And the really disruptive thing, besides the first fact that someone had these smartphones, was that companies like Uber, Lyft and others went and owned sort of the end to end experience to make it possible.
- [31:44]
A
Yeah, I think it's this sort of all or nothing thing where Uber and Uber, like services, they kind of only work as well as they do. I mean, okay, they've all got their problems in specific areas, but let's just say like most people would agree that having these services is net positive now over just having taxis available, but the only kind of work, because it was all or nothing. It wasn't, oh, let's try and adapt taxi services into this. It was, we're just going to start kind of from scratch. It's going to take a ton of money, ton of investment. And the tipping point, so to speak, was that so many people then had a smartphone with acceptable connection in their pocket, which means it's the same for the driver as it is for the person asking to have the car. The driver needs to have this connectivity and technology as well as the person asking for it.
- [32:35]
B
Yeah, and I think people forget about probably all the money that Uber burned to get to the economies of scale. Because in the early days with companies like Uber and Lyft, they used to have to subsidize drivers to just have them stay up there because they didn't have enough riders essentially to always be requesting. But if you were in the early days of say Uber, you download the app, you spin it up and there's no cars available, you're gone, you'll never use that thing again. They had to essentially pay drivers to be available so people could have a good experience because they're building a two sided marketplace. It's the classic chicken and egg problem. So you have to subsidize one side of the market and lose money in order to build both sides of the equation and balance it eventually. But that Takes a tremendous amount of investment and years of laying the foundations in order to get something where you have these network effects and it starts to grow on its own and you start to basically print money.
- [33:29]
A
Yeah. Then maybe leads quite nicely into another type of technology bets, which is more like the moonshot stuff. When we say moonshot, we mean something that really does sound far out from a movie set 100 years in the future. And when autonomous driving was starting to be looked into, that's kind of what it really felt like. This is crazy. Having cars or vehicles just generally on public roads that do not require a human at all. That just sounded crazy. And here we are, 2023, I think, was Waymo coming on the roads. I could be wrong about that year. But certainly in the last couple of years, I think is when it's actually publicly available, that you can open up your phone and just call a Waymo and this car pitches up, which is pretty exceptional. And again, I think now that people get to experience that technology, I think the consensus is, again, it's net positive. Okay. There's a whole bunch of debate around the economics and kind of removing human jobs of drivers and so on. But the net positive actually is around people say, I'd like to get in a car and not have to think who this driver, this human is. I don't want to talk to this person. I don't want to hear their music. I just want to get in my car, be driven somewhere as if it's my car, and not have to worry about anything else. I can identify with that. The States is somewhere where you get into an Uber and suddenly you're being talked to a lot, which is not what I'm used to. When you're hearing about this person's political views and all sorts of things, and then you should have to say, hey, buddy, can I just put my earbuds and had a long flight.
- [35:10]
B
Yeah. I think that everybody who's taken any kind of rideshare has had their fair share of probably uncomfortable situations in those rideshares. And then I think, in particular, I have kids. They're not at the age where I would shove them into Waymo or something like that, but at some point when they're older and they're involved in school activities and stuff, and I know parents that do this with some of their older kids, if they did have to take a car somewhere, I feel a lot more comfortable putting them in a robot car where no driver is going to put them in an uncomfortable situation. And I could track them on My phone and I know it's locked, I know they've arrived there and all this type of stuff that feels essentially a lot safer than having them with some stranger driving them around. So I think there is a lot of net positive and this is a good example of something that took decades essentially investment and we're still not to massive, massive scale. But these are available in a number of cities here in the US now and people are experimenting in a lot of different places. So the technology is there and it's something where it has an extremely high bar when it comes to safety and there's a lot of regulations to navigate as well, which is probably in part what has taken as long as it has. But that's a very, very complex issue. So I think when people are all debating is AI real or not? It's like we'll walk around the streets of San Francisco and see a car driving itself and picking up a driver. Yes, that is amazing. It goes back to this argument we were talking about earlier of fake it until you make it. It's like yes, there is some snake oil stuff out there, but there is some real value, 100% behind some of this technology.
- [36:55]
A
Exactly. Thinking about what the actual tipping point with autonomous driving was, as you say, regulation was a big one. Technology was obviously a huge one. It wasn't that of creating an autonomous car is straightforward. I think they quite quickly. I mean there's been an autonomous car driving competition for years and years and years. That's actually I believe where the main person at Waymo or there was the self driving truck that had sort of slight legal issues with Google. The founder of that, he was often winning these competitions but these are still in closed environments, closed situations. There's a whole different ballgame when you put an autonomous car amongst other non autonomous vehicles and having to follow roads and so on and so forth. So yeah, the tipping point seem to be regulation technology and quite frankly it does actually need people to want to use this thing. So I think that's interesting that. I think Waymo's done quite a good job of making the experience such that people do actually want to choose these over a traditional taxi or ride hailing if it's available.
- [38:06]
B
Yeah, I mean I think they can really optimize for the rider experience. They can essentially train the vehicle to drive in a way that's comfortable for the rider. Whereas if you're driving with an Uber driver or whatever service that you're using, there's not necessarily other than the star system, but there's not Necessarily incentive for the. Or the driver might not even know that. Maybe they're just slamming on the brakes constantly and you're rocking around in the back starting to get sick. And it's not like they're doing that necessarily on purpose. That's just the way they drive. While the robot car doesn't need to operate that way if they're sort of trying to optimize for comfort and having a good ride and the rider experience.
- [38:46]
A
Yeah, the consistency of the service is huge. I agree. Getting in any ride share of service, I'm just always dreading if it's one of these sort of stop start drivers. So yeah, I mean as we're kind of figuring out here, tipping points have quite a different reason for many different technologies. I think this is where people maybe get a little bit confused as to sort of why is there investment going in here, there and everywhere? It will never work. Well, there's just so many conditions under which a technology will make its case and people actually start adopting it. That kind of brings us on to AI and I think looking at that versus the Internet is kind of interesting. They both had quite similar, I guess, investment profiles like huge amounts of infrastructure investment was needed. And in Internet it was actual laying of cables. For AI, it is data centers production of chips which there are shortages of. But yeah, I think you've got some kind of quite nice ways to look at this, Sean, in terms of Internet versus AI.
- [39:47]
B
Yeah, we touched on some of this stuff with some of these other technologies and tipping points where I think there's just different risk factors. With the Internet, a lot of the risk was an economic risk. No one knew how to make money from it and whether it would even happen. And then they had to spend years laying some of the physical cable and also wait for enough people to be on the Internet for it to have mass appeal. So I think in the late 90s there was this risk of essentially is this useful. And with. With AI, I think the risk or the question that people have is more about is it too expensive? Which ideally is. That's a better problem to have because I think that we've seen with other technology, other types of investments, with economies of scale, the expense has typically gone down. And the other thing with the Internet was there's this big question about the behavioral change. Would people actually stop going to the mall to buy things versus buying those things online? I think mobile had a similar thing where people thought no one's ever going to book airline tickets from a phone or with the company that I started and built. The big question there. No one believed that anyone would ever apply for a job on a mobile phone. We knew that that wasn't the case because we saw people doing it in convoluted ways in the early days of mobile devices. Now tons and tons of people do that but at the time it was this unknown thing of is this behavioral change going to happen or not. And I think with AI there isn't this question of economic and there isn't necessarily a question of behavioral change. If you can make the technology work, if you can make AI work I think there's clear economic value. If AI writes the code, it answers support tickets, it designs drugs, it diagnoses diseases. That's not really a question of whether it's useful or not like there was with the Internet.
- [41:43]
A
I completely agree. Yeah. When people, and these are generally non technology people, as I was saying at the beginning, are asking me is there a bubble? And I'm saying well that's the financial question. I just keep saying I think there's just such a long way to go for a lot of when I say non tech people I just mean people not working really day to day in the weeds of tech. Understanding how AI is going to help them. I gave to this person who's a few years older than me, sort of closer to my parents age perhaps than me even. I did actually give the example of Claude has completely removed me from needing to be my parents tech support because I live quite far away from them. But I would still to this day get Messages, hey, my MacBook has done this thing, what do I do? And usually I just sort of look up Apple support and then I just ping them the link on that. Nowadays I literally just say Claude will have a better answer than me, please use Claude. And and it works. And they're really sort of, they're in their 70s and they are just amazed by this. And to them. So it's really incredible once someone is shown the technology and how far it can take them. But I think there must be so many people in the world right now that just unfortunately don't have access to this for many reasons. Or they do have access or could have access but just it feels scary and they don't want to touch it yet. It could help them in so many ways.
- [43:05]
B
Yeah, I think there's always more resistance to technology that feels somehow human or it's mimicking something that's historically been associated as a uniquely human skill. I think driving is the same thing. You have these robot cars that feels Threatening. If you actually have robots that look humanoid, I think that feels threatening. And then certainly AI that can generate content or speak or create an image or something like that historically has been associated with human skills. Also feels uniquely threatening.
- [43:39]
A
Yeah. So I guess just to wrap up on this, on what is a tipping point generally. I mean, infrastructure seems to underpin a lot of it. A technology can do better when the infrastructure is kind of there. We saw that with the Internet. Once enough cables had been laid, enough routers around the world had been positioned, and everyone could talk to everybody. That's the kind of tipping point when Internet gets its mass adoption not just in the United States, it explodes everywhere kind of the same. When we looked at the autonomous driving and that, again, if you want to look at infrastructure, well, okay, the roads were there, but the infrastructure within the regulation wasn't there, for example. And technology, you could say as infrastructure, the computing power needed to figure out how Waymo drives around. Yeah. Wasn't there for a long time. And then especially, how do you put that technology inside a car? It could sit inside a giant data center maybe, but inside a car. Okay, well then you need really fast or not inside the car. You need very fast speeds of data. We didn't have 4G or 5G until very recently. So infrastructure seems to drive a lot of this. And that's probably what we're seeing with AI as well.
- [44:53]
B
Yeah, I've been saying that for a while, that whenever you have these kind of technology shifts, the first couple years is really laying the foundation of the infrastructure. And then it takes. Once you've done that, then companies or people figure out, how do I create this uniquely mobile experience? It took years before we had Uber and Instagram and these sort of mobile first experiences that people started using and everyone's familiar with now, but in the early days of mobile, it wasn't like that. And it's just like the same thing where it took a decade of laying the foundations of the Internet before you had things like Google and mass economies being created through the Internet.
- [45:34]
A
Yeah. Okay, so hopefully that's been helpful. Just hearing and thinking about tipping points, maybe for something you're working on right now, you're thinking, this will never work, this will never get adopted. Well, maybe just sort of think about all the pieces needed to get there. Moving on to what we tend to think of as our favorite part of the show, hacker news highlights. So Sean and I just pull out anything that's maybe caught our eye in the last couple of weeks. Do you want to kick us Off Shaun.
- [46:02]
B
Yeah, so usually I end up choosing some obscure hacky project that someone did. So this one's a little bit more mainstream, but I thought it was so interesting I just couldn't help but call it out, which was this article about how, how scientists were able to put moss on the outside of the International Space Station and the moss survived for nine months and then they kept it growing back on Earth. So the article goes into details about all the different types of moss they experimented with and how a lot of it continued to grow essentially in this vacuum of space, pretty unaffected. I think there was some stuff where if it was in direct sunlight for too long then that that was a problem. But for the most part moss can survive in space, which seems insane to me.
- [46:48]
A
Yeah, that's pretty insane. So yeah, thanks user Geox, I think it was who posted that one. Yeah, I'm definitely taking, I guess a holiday light hearted approach to hacker news highlights this month. So not going any super deep end developer stuff. I think it's nice just to look at something else. So I've got a couple of kind of more design related ones actually. The first is, it's called Fran Sans, a font inspired by San Francisco light rail displays. This was so interesting. Basically someone who designs fonts, but they wanted to emulate the font that the light rail in SF uses. So these are kind of like tram, like kind of trains. I believe I've written them before. I'm just sort of trying to think how to describe them to anyone that's not been in sf. But I think tram is sort of like a good description. And it's the fact that these displays, they're not like LCD displays, they're kind of these big boxes that have these shapes that become the letters, but it is a font to itself. And then it's like the glow of the light that comes through the box that kind of gives the font its kind of personality. So the author of this article actually got to go and hang out where they make these boxes and they got to play around with them and then yeah, this font has now been produced and they showed how it's now been used on some quite cool theater posters for I think for London Shakespeare Theatre Company, et cetera. I just love this stuff. I guess I started off really as a front end developer and fonts was always something I just loved, so spending far too much time on. So yeah, this kind of caught my eye just because I think it's sometimes nice to get lost in something that does apply to technology. It's A font that you can download and use in your projects, but equally just sort of seeing the craft behind it was awesome.
- [48:46]
B
Yeah, it's really cool. Yeah. And then the other one I pulled was about. It caught my eye, but it partly caught my eye because I have a lot of conversations with businesses in the industry and of course, one of the topics of conversation is a lot of times security privacy around AI. How are you as a company investing in this technology and how do you gate access to certain things? What are the principles that you create, just like you would for businesses, get vendor procurement and have certain security expectations when people are trying to adapt that to AI. So this article was the general principles of using AI as cern, and they lay out nine different principles of that. And I just thought it was a useful reference. So a couple of examples are like, every investment needs to address transparency and explainability, lawfulness and conduct sustainability. Human oversight is a big part of it, data privacy and so forth. So I thought it was a good reference. It's something that comes up a lot, I think, as a topic of conversation.
- [49:49]
A
Yeah, no, it's good. It's good to see that kind of thing appearing on Hacker News as well. So, yeah, like things that sort of actually drive change, but also drive responsibility. So. Yeah, I like that. Yeah. I think the other one, this got voted up so far, further than I would have expected, maybe for something like this. But it was called the Toy Story youy Remember. And this was basically just. I think quite a lot of us maybe listening are of an age where Toy Story was. You know, we were kids when that film came out. And the sort of TLDR is that. That Pixar's films, well, they used to literally be on 35 mil film. And that's the difference between kind of the look of the film versus what, you know, they've all been like, fully digitally converted. So, you know, when you're watching it, you're not watching the 35 mil on a DVD even. It's like it's all been digitally converted. So it's a really good article because it's got so many comparison photos. It shows like what toy story on 35 mil looked like, and then now what it looks like on Disney. And just the kind of aesthetic that comes through. If I look at the. There's a top image of the 35 mil. That, to me is Toy Story, as I think about it. And there was just something about the colors and the slight softness to the whole thing which you don't get with these high definition Digital equivalents.
- [51:15]
B
Yeah. Some of these are pretty stark contrasts. Like if you look at the Aladdin 135 mil versus Blu Ray or the Lion King is another one where it's 35 mil versus Blu Ray.
- [51:26]
A
Lion King. Yeah, that really. I mean, that's again a film that I watched as a very young child. And yeah, it's like, wow. To me the 35 mil version just looks so much nicer because it's sort of got so much more personality to it.
- [51:38]
B
Yeah, absolutely.
- [51:40]
A
Yeah, I thought that was super interesting.
- [51:42]
B
And related to that, I saw something recently about an article talking about why do movies feel less grounded in reality now? And it was kind of comparing how they would film people say from the 90s in a movie versus now. Everything. Like the backgrounds are always sort of soft in order to put the actor in the foreground. But the way you see in reality isn't like that. If you look at a person face to face in reality, it's not like their background is faded out and their face is prominent. So by filming that way, they're kind of creating this false reality that feels off. It's essentially the argument of the article.
- [52:20]
A
Yeah, I mean, as a lot of people have, I've dabbled in photography. I've got a full frame camera that I love to pull out maybe once a year. But yeah, I felt like there's been this phase towards using. It's called the F stop. Like using very low F numbers basically. Which basically means. Yeah, you focus only on what's in front of you. And like this is, I think because these cameras can do film now and they're like so many people have access to them now. Like so much of what we watch these days, it just kind of is a. I think it's a bit of a hack. It's like a bit of a hack to make a film look really cool and professional and nice is just dial down the F stop. Suddenly things are in focus and then they pass in and out of focus and that looks really cool. And that's what you say, Sean. It's like it isn't reality, but it just looks and feels nice.
- [53:08]
B
Yeah.
- [53:09]
A
Okay, well, thank you again for tuning in to SED News. I hope that if you celebrate holidays at this time of year, everyone has a good respective holiday. Sean's off to have his American Thanksgiving after this. So thanks for joining us. Just before that, Sean. And yeah, we'll probably catch everyone again, I guess, after Christmas time and try and wrap up what's been going on. I don't think predictions are even going to. Who knows, quite frankly, what's going to happen across the holiday season? Because things do dial down. So let's just come back and see. Almost like a present in itself. Like, let's just see what's happened when we come back.
- [53:44]
B
Yeah, sounds good. Thanks, everyone.
- [53:46]
A
Thanks a lot.