Ask Paul: February 21 (Premium)

Happy Friday! AI angst is understandably on the rise, and we look backwards and forwards as we cruise into a much-needed weekend.

?‍?‍? They think they’re people

oasis21 asks:

What’s the argument for AI replacing people? I wanted to hear your thoughts on this.

Generally speaking, it’s the same rationale as any technology replacing people. And that usually comes down to cost, once you get past the efficacy of the solution. Anyone can probably come up with dozens of examples where some series of cost-cutting steps led to outsourcing and then automation, perhaps with several interim steps. For example, a support call center being national instead of local (as with an Internet provider or phone company) and then being moved out of the United States (to India, perhaps, or whatever lower-cost international location). Between that and full AI-based automation, we’ve seen various solutions to save costs there further, such as the phone-based (or web-based) menus that step you through self-help flowcharts to fix whatever the issue is without engaging a human being at all.

AI (particularly ChatBots) have made me more productive and I find them useful. But I don’t see how a chatbot can replace a human being. I see this AI stuff as a tool.

Well, this is very specific to the solution, need, or whatever. But a chatbot, literally, as we know and use them today, is unlikely to replace a human doing a very specific job. That said, I happened to use a support call center as an example above, and it’s not difficult to imagine AI leading to those companies needing fewer humans. Currently, most of those interactions are annoying and flowchart-based, with the person on the other end of the line often not really knowing the answer themselves anyway. If a company like that can use AI to speed time to answer, and make that business more efficient, then it will be able to provide that service with fewer people. And that will save it money.

We live in the Lehigh Valley in Pennsylvania, and there is a rich history there of it being a nexus of transportation, which became very important when the steel industry took off and became a big part of the national economy. But getting people or things from here to there went through multiple stages of overlapping evolution, from horses and stage coaches, to boats (rivers, but also manmade canals of all kinds), to trains, to cars, to trucks, to the national highway system, and to planes. And now we have advances like electric vehicles, drones, and whatever else. And today, most of these things are still in use in some way (maybe not horses), and logistics and shipping companies can use technology–increasingly AI–to figure out the best/most efficient ways to get people and things from here to there.

Is there some next loss of jobs in this story somewhere? Well, yes. But it’s also just part of the story because that allows humans to do other work that is either less rote or physical or whatever it is and is perhaps more rewarding. The goal of any technological improvement is … improvement.

While being conspiratorial, AI might be an excuse to layoff employees and save money, while the remaining employees just get more responsibilities and are in fear of loosing their jobs. Is the argument that workers are more productive with AI, so companies can lay people off? Even then, if a worker is 5% productive, and you gave 20 employees access to AI, I still don’t see how that can replace an employee. This whole thing assumes that the responsibilities of an employee can be broken down like that to being with. To me, it feels like hiring nine women to give birth to a baby in one month.

AI will absolutely be used as an excuse to lay off employees, but that doesn’t mean that’s how it will be used generally. Companies always try to find excuses for any cost saving exercise that leads to bad news for employees. And companies have always laid off employees. They don’t need AI to do that.

But whatever. Again, this is all very specific to whatever jobs and situations.

Here in Mexico, the gas stations are full service, and this reminded me of a thing I barely experienced as a young child but anyone can sort of enjoy through a scene in Back to Future where a car would pull into a gas station and four or five guys in immaculate mechanic jumpsuits would run over, each doing something for the driver: Filling the gas, of course, but also checking the oil, cleaning the windows, checking the air pressure in the tires, and so on. This is fanciful today in the US, with a few exceptions (New Jersey, maybe Oregon), you can’t even find a full-service gas station, and when you do, it’s just somebody pumping the gas. These companies didn’t need AI to eliminate most of the jobs that used to be part of that experience. And we as customers think nothing of it. It’s been a few decades. And we see this self-help/self-checkout experience everywhere. Supermarkets, fast food restaurants, whatever. It’s always about saving money.

Given that, AI is not an excuse to layoff people so much as it is a reason to reevaluate where people are needed and how many. Again, very specific to each use case. But generally, AI will enable businesses to provide whatever services more efficiently and will save them money. And yes, that will lead to fewer jobs for humans in areas that previously required them. This is, if anything, normal. I’m sure there were frictions in every single transition like this. Boat companies that did what they could do to halt the expansion of trains, or whatever. That, too, is normal. That’s my Exchange administrator story in a nutshell: The old guard resisting change because it impacts their jobs, even though it’s the right thing for the company for cost and efficiency reasons.

Although, there are edge cases. AI is pretty good at summarizing, so I can see an employer laying off a person if that was their only job. There’s probably more examples out there too. I will admit, I don’t know everything — I could be wrong.

Well. Look, diminishing AI as just being good at summarizing is sort of missing the forest for the trees. What you’re describing is very specifically a subset of generative AI which is itself a subset of AI and machine learning. And we’re ignoring how fast this is improving.

So let me be specific.

I’m a writer. And I’m also a certain kind of writer, with a specific focus and whatever point of view. There is nothing that I do that cannot be replaced by AI in time. Much of it right now. Eventually, all of it. What I offer, what my wife and I offer at Eternal Spring, is really the net result of decades of experience, what I think of as strong opinions based on that experience, and a desire to help people not make mistakes we’d maybe made in the past. But even perspective like that can be learned by AI. It’s going to happen quick.

In my professional lifetime, we’ve moved from paper-based books and (enormous) magazines to whatever range of digital publishing capabilities. There were ups and downs, but from a business perspective, it was always about reaching as many people as possible while saving as much money as possible. Paper-based publishing is no longer mainstream or even desired by most. But all the great publications of the past are either dead or nothing like their former selves. PC World translates (with AI, I bet) and republishes horrible quality posts from European blogs. Neowin hires “writers” who are so inexpensive and inexperienced that their output seems more like English as a third language than English as a second language. And the next step for both, and for so many publications, will just be AI. Those people will all be out of jobs.

You can do this yourself, right now. Find a blog post or press release from whatever company. Copy it to the clipboard, feed it to any AI, and ask it to rewrite it. The results are usually–not always, but increasingly so–quite good. And I want to be super-clear about this. Though I try to always provide some perspective in whatever I write, this is exactly the work that bloggers have been doing for 20 years: Rewriting news from an original source, often badly. Ignoring for a moment that professional writers, reporters, or press of whatever kind looked down on bloggers (and many still do), they’re all about to be replaced. All of them. All of us.

Are those job losses regrettable? Maybe today. But as I told my wife recently, you can imagine some future generation, maybe my grandkids, wondering why on earth anyone would have paid us to write words about a topic that will then be automated and can happen in a millisecond with AI. Just like we now wonder why women used to sell cigarettes and tea and whatever else on trays on airplanes. These are jobs that made sense until they didn’t.

AI will generate fictional books better, faster, and more efficiently than people. It will win awards. It will generate TV show plots and movies and whatever other content. Video games, obviously. Imagine a Call of Duty where every deathmatch multiplayer level is different and not just the same six tired maps over and over again, and each is perfect from a playability perspective because the AI knows what works, what gamers like, and has an infinite capacity to make these things on the fly. This is objectively “better” than the situation today. But for now, in the short term, we will worry about job losses in the video game industry. Long term, we’ll wonder why we ever needed those jobs in the first place. Why could all those humans only make six levels after a year of work and at great expense?

This is what always happens with technology advances. But AI is happening quickly, which makes it confusing and scary. And its reach is broad. It will literally be everywhere. And this is why I’ve moved past the hand wringing and wondering and have accepted this as inevitable. Everyone will get there at their own pace, and have that moment where it finally makes sense to them. But the world isn’t waiting. These efficiency gains are going to happen everywhere, whether the people impacted see it coming or not.

Anyway. I feel like there’s no real argument here. AI will replace people (jobs). I’m sure it’s doing so already. Short term, this is terrible for those people. Long term, it will be better overall.

Tied to this, train_wreck asks:

How should society handle job losses due to AI, and how do you actually see it panning out? If history is anything to go by, it will be handled by essentially giving the middle finger to those people who are made redundant. This troubles me greatly, and in my ideal world we charge ahead with AI/automation while at the same time offering abundant resources for reskilling. (But of course we won’t do that. See: horse carriage manufacturers, portrait painters, coal miners, etc.)

The problem for us here in 2025 is that we’re living through this in real time. As noted, it’s happening quickly, and we don’t have much time to understand what’s happening, let alone respond to it effectively. You can be cynical about it. But when it comes to job losses, it’s important to realize that these arefor-profit companies, not charities that exist to benefit mankind. They’re going to cut costs where they can, and human costs are always the greatest. Per the outsourcing example above, jobs will move to lower cost locales–nationally and then internationally–and then even those jobs will be too expensive.

This is why the notion of building microprocessors and other chips in the US is so ludicrous in some ways: The typical cost of an employee in the US, between salary, healthcare, and other benefits, is exponentially higher than it is in China, Vietnam, or India, and I’m sure we’ll move from there to Africa next. So why do it? Because it can be automated and we won’t need as many human jobs. Plus subsidies and whatever other national/political reasons. One might argue that AI is what makes US-based chip making viable again.

I often think of the “advice” that was given to Appalachian coal miners who were suddenly out of jobs: Just learn computer programming. Yeah, right. Like there is anything in their history, generally, to suggest that this is the next logical step. But whatever. There’s no good answer there. Who pays to retrain these people? We’re a country built on individualism, so there’s no social safety net for anyone, that’s on them. I don’t even like paying taxes for a school I have no children in, why would I or anyone else want to pay for that? It’s a problem. But this is a problem that just keeps repeating itself. It’s a today problem.

I mentioned the Lehigh Valley above. That steel industry is long gone, as are whatever jobs there were. This went from being the center of the economy to being nothing. And there are places like Bethlehem or Pittsburgh (in western PA) that suffered for quite a while because of that shift. Both have bounced back nicely. As did the former mill/factory towns around Boston. As has Detroit to some degree, after losing all the auto industry jobs. There are former ghost towns near Albuquerque, New Mexico that were so frozen in time they were perfect for movies like Young Guns, but they turned into little artist communities and viable little towns again. It’s never perfect. But there are ups and downs and we move on.

Morally, I’m with you. But we’re not changing this country. We’re in an era of stupidity and misinformation, and we demonize terms like socialism that would benefit us as people. It’s too bad.

? Origamic

jgraebner asks:

The discussion of Courier last week brought to mind Microsoft’s somewhat similar “Origami” Ultra-Mobile PC (UMPC) project that did make it to market, although it was short-lived and never attracted any major manufacturers other than Samsung. I was wondering if you had any insight into it. I actually did buy one (a TabletKiosk eo) back in the day and, in retrospect, I think it was an interesting pre-cursor to the Surface Pro where the idea was ahead of what the available technology of the time could reasonably support. It was a fun gadget and a conversation starter when I used it in public places, but it definitely wasn’t very usable.

I did own a few UMPCs, including the Samsung you mention. This was an interesting era, especially for those in the Microsoft space. The world was moving on, but Microsoft didn’t yet get the memo, and it threw a lot of proverbial spaghetti on the wall, trying to find consumer platforms that could replicate the success it had experienced with Windows and then emulated with Office and Server. But nothing worked.

One of the big debates in this time period–I’ll frame it as roughly the 2000s, from the advent of MP3 players through the introduction of the modern smartphone era–was how to go to market. Microsoft was so Windows-focused that using that model seemed like the right choice, and it always tried to protect Windows. And so things like Media Center would be complex PCs instead of Xbox-like devices. (Though, remember that the first Xbox was a PC, basically) Tablet PC was a PC, not a Windows CE-based device. And UMPC was … a PC. Not a device. It was the wrong choice. But it was where Microsoft’s head was at in this era. Whatever the problem, Windows was the answer. Until it wasn’t. Those shifts are difficult and, to be fair to Microsoft, many companies never change. It did at least move on, in time.

Someday AI will do this for me–someday soon, come to think of it–but manually looking through my archives, I see that Samsung released the first UMPC in March 2006 and that I was told in a briefing that iteration on the platform was “fast and furious,” and that Microsoft was “looking for opportunities between the phone and typical laptops.” Part of the Windows focus at Microsoft was about partnering, too, that was part of the success there, and so it worked with OEMs to see which would come onboard. In this case, Samsung was first. The software dated back to the Windows XP Touchpack, which I can’t say I remember, and Microsoft defined this category of devices as being 7-inch displays or smaller and weighing 3 pounds or less, and—wait for it–running Windows. This was all inspired in part by the relative success of the OQO mini portable PC you may remember.

Taking a step back from the meeting notes I’m quoting there, this was 2006, 2007. Apple was just about to upend the smartphone market and personal technology with it. We had finally transitioned to Vista after all the Longhorn terribleness. This was just one of so many consumer software/hardware initiatives, some of which came and went in a flash. I guess UMPC is in that list.

Anyway, Microsoft assumed that consumers and dual use (work/home) customers would be interested, what they called mobile pros. The value was using the full PC experience in more places, with familiar experiences and apps (and given how Apple went in a completely different direction, you can see now this was a mistake). Battery life was challenging, but the smaller screen helped a bit. The first version used a modified XP Tablet PC version of Windows. But then it was optimized and Microsoft moved to Vista and the Origami experience. Finger-based navigation and on-screen selection. ASUS, Fujitsu, Medion, OQO, Founder (in China), and Sony were onboard. HTC. Intel came out with the Centrino chipset, and Via became an option. There were logo requirements, of course there were. Oh, Microsoft.

The software was interesting. It had a crescent-shaped virtual keyboard called DialKeys. Full-screen and touch-based media experiences. You could play games, allegedly (Sudoku was included). There were media player controls on every screen (I was told, don’t recall that). “This is all about consumption,” Microsoft told me in what might be my first exposure to that term. There was a Microsoft Reader app for UMPC.

What a weird era this was. These devices were moving from hard drives to solid state storage. From 800 x 480 resistive touch screens to 1024 x 600 integrated touch panels. From 512 MB to 1 GB of RAM to, well, more RAM. To passive cooling and lower-power chips. From 2 to 3 hours of battery to 3 to 4 hours of battery.

This warrants some kind of write-up at some point, I guess. I held onto at least one UMPC for many, many years, but it’s now long gone due to whichever decluttering push. I think it made it all the way to Pennsylvania, so 2017 or so. Crazy.

But for now, I’ll just summarize this with Microsoft went to market the way it did at the time. And Apple showed it, and the world, that it was possible and better to be choosy about form factors, customize the software platform for each instead of just building more on top of the PC-class OS, and build these things as devices not mini-computers. The partner vs. monolithic unilateralism thing is still debatable, but when you give choice via partners, you also concede on quality and consistency. And Microsoft has a history of over-engineering things, with big, heavy platforms. Apple was always more targeted, more lightweight, and quicker. That model won, regardless.

Surface came about because of many years of frustration on Microsoft’s part with its hardware partners. They were more interested in lowering prices and improving margins than they were with quality or with aligning on Microsoft’s strategies. When Apple came out with the MacBook Air, it wanted thin and light PCs like that, but there were no takers. When hardware makers from Asia came out with tiny laptops running Linux, chipmakers and PC makers embraced that model and created the netbook, forever destroying their bottom lines. For better or worse, Microsoft wanted to move the industry to more of a device-based, high-quality future. But it had no support. So it did the unthinkable. I can’t say that worked out, but typing this on a Surface Laptop 7, I can say that it took a long time, but we’re finally in a good place with a device-type platform and high quality.

Surface Duo wasn’t successful, but there are interesting comparisons to make with that an UMPC because they both came out of different eras with different go to market strategies. The partner bit was never going to happen. Basing it on Windows didn’t make sense. For whatever its flaws, Surface Duo was at least true to the form factor. UMPC was … what everything Microsoft did then was. It was kind of a mess. A glorious mess, in some ways. But also a slice in time when the ship had sailed and Microsoft didn’t yet realize it.

? GPUs in the mist

jrzoomer asks:

Paul what are your thoughts on the new NVIDIA gaming GPUs the RTX 5090, 5080 and 5070 Ti, and specifically the frustrating lack of supply in the retail channels? Many of these are going for minimum $1400 on the 5070 Ti up to $6000+ for the 5090 on places like eBay.

I have to think that Nvidia is much more concerned with the datacenter/cloud AI market than gaming GPUs at this point. And that its supply issues are tied to that. Looking at its most recent quarter, Nvidia’s Data Center business earned about $31 billion, compared to $3.3 billion from gaming and AI PC. 83 percent of Nvidia’s revenues come from Data Center now. And the growth numbers are just as telling: 112 percent for Data Center and 15 percent for gaming and AI PC. (AMD’s revenues have similarly shifted away from graphics cards for gaming.)

My thesis is that the recent uptick in PS5 console sales this quarter is related to this, ie PC gamers jumping ship into consoles, given the extraordinary cost and flat out inability to obtain these new GPUs. What do you think?

I have a hard time understanding the PS5 uptick. It can’t have been from PS5 Pro, given that it’s so expensive and Sony would have absolutely called it out if there was even a shred of evidence to support that claim. It’s not like this was a big quarter for splashy new game releases or whatever. It’s very strange.

It will take a while before we know whether there’s any trend in PC gaming and whether that’s related to the PS5, but I feel like that market is pretty consistent. And as I’ve been writing recently, now that integrated graphics are good enough for most gamers–seriously, the latest-generation AMD-based PCs are off the charts–I could imagine that market growing overall. Maybe not so much for standalone graphics cards, but PC gaming in general.

But I really don’t know. I’m not sure the PS5 scratches the same itch as a gaming PC. I can’t imagine this had any/much effect, but perhaps Xbox coming through on its cross-platform promises post-acquisition helped a little bit. The latest Call of Duty is obviously off to a great start, and most of that is on PlayStation. Is the PS5 the best place to play that game? Probably.

? Microsoft Launcher

Ruvger asks:

What’s happened to the Microsoft Android Launcher? I see its in the store but I haven’t heard anything about it for ages. Is it still being developed?

I don’t know, but the silence from that group is perhaps telling. I can’t imagine there’s a lot of interest in such a thing, and that third-party Android launchers are a small but dedicated market overall. But also a free product.

Looking at the Play Store listing, you can see it was last updated in January, but that update is likely just a few bug fixes and the like. The website that listing links to is a page on Microsoft Support, which is a bad sign, but there is a better (though dated) landing page on the Microsoft Garage site. (That hilariously has a quote from me, I see.) But I can’t find anything recent about this thing. It used to be more prominent in the Microsoft website.

So I will guess.

I can’t imagine this thing is an ongoing concern, and that its last hurrah was almost certainly whatever work it contributed to the Surface Duo line. And that with that product’s death, there was no real place for a formal push to create a Microsoft-oriented Android experience. This is too bad on whatever level, but it may also be a mistake. I’ve long felt that there is a market, if only from businesses and maybe some enthusiasts, for Microsoft 365-focused Android devices. And that a Surface Pro-like tablet running Android and this launcher could find a reasonably sized audience. I had assumed that Duo would be the first of several Android devices, but perhaps it did even worse than expected and that was that.

So perhaps the plan now is to simply ensure that its mobile app experiences are as good as they can be and not worry about a dedicated launcher. I have to further assume that the curious Samsung partnership is something Microsoft would like to duplicate with other hardware makers, so it can get its technologies more seamlessly integrated into more devices. (Though just having the market leader doesn’t hurt.) But yeah, sorry. I can only guess.

(All this said, Microsoft did integrate Copilot into Launcher last year.)

Speaking of failing on Android…

? Amazon and Android, sitting in a tree, ignoring each other completely

helix2301 asks:

I saw my post get taken down from the forums. I just wanted to know your take on Amazon getting rid of the Android store, which we knew was coming. Might be a big hit for people still on the older hardware. Leo brought up that Kindle is Android under the hood. Just wondering about your take on that and their new OS.

Sorry, when a forum post covers a topic we wrote about, I redirect it to our news post so there aren’t two sets of comments/discussions.

As for the Amazon Appstore for Android news, this one was confusing to me before I understood that it wasn’t tied to Amazon’s devices. What it is retiring is the standalone Amazon Appstore app for Android that users on non-Amazon Android devices would need to download and use alongside Google Play Store. And it’s clear few did so. Why would they?

One thing I do wonder about is whether this is tied in any way to the news from a year ago about Microsoft removing the Windows Subsystem for Android and its Amazon Appstore from Windows 11. That is, one might assume that that played some role in Amazon’s announcement. But I wonder if the reverse is true. Perhaps Amazon has been planning this for some time, alerted Microsoft, and they just acted first. It’s impossible to know.

In any event, Amazon will continue to use its lackluster Appstore on Fire tables and Fire TV devices. This might be a sunk costs thing, but Amazon is so cheap that paying for what everyone really wants–the Google Play Store–is apparently too expensive for it. But Amazon’s devices could be a lot more compelling than they are. I wish they would rethink that.

As for the Kindle e-book readers, I had assumed they were AOSP (Android) under the hood, just like the Kindle Fire tablets, but using even less powerful processors and with minimal system resources. But Googling this, it looks like it’s an even more stripped down Linux version of some kind. I don’t think there’s much in the way of new there, beyond the recently introduced color screen option, but I’ve always found the performance pokey and the screen refresh/flip less than ideal. The Colorsoft uses a MediaTek MT8113 processor, which apparently is e-reader specific (and uses a 12-nm manufacturing process!)

I do know that Amazon used to write the Kindle e-reader user interface to Java, but it switched to the web-based React Native in the past year or so. This probably makes sense for this kind of device.

Chime being shut down is not a big deal I think that was never a success. I do think it’s interesting that Libra office getting online collaboration peace like SharePoint. I think these open-source alternatives are great and have a place but the lack of enterprise-level features is their downfall in business space. I don’t know any businesses not using 365 or Gsuite.

Yeah, I feel like this market has settled into a natural end point with two major competitors. But the interesting thing about Google Workspace is that those types of companies seem more likely to use any number of third-party tools as well–things like Notion, Slack, Zoom, and so on–where Microsoft shops, which tend to be bigger companies, are more likely to stick with what they pay for, for the most part. It will be interesting to see whether this shifts over time as smaller companies grow bigger. But I don’t see another major Microsoft 365 competitor to ever threaten them. It will be death by a thousand cuts instead. If at all: These things can coexist as well.