Ask Paul: April 25 (Premium)

Happy Friday! I got off lightly last week, but we’re making up for that now with one of the longer Ask Paul’s in recent memory. Let’s kick it off…

Note: We were in Oaxaca over the long weekend, so I threw in some photos from that side-trip.

? Trusting AI

wright_is asks:

Do you think we will reach a stage where we can actually trust LLMs?

Apologies for putting the end of your question first. But to answer that one, yes, most of us already do.

And I don’t mean that cynically. I still struggle to use AI in my day-to-day work, which is a cute way of saying, I still don’t ever use AI in my day-to-day work. But since Microsoft made GitHub Copilot free in late 2024 (with limits I’ve never run into), and with the rise of alternate tools like Cursor and the built-in coding capabilities in all mainstream AI models, I have started using it for my current coding project. And I discovered something interesting. Maybe two somethings.

First, the AI is so good that I just started trusting it implicitly. In the latest article in the series I’m writing about that work, I mentioned one example of how I use this almost in passing, but GitHub Copilot (in that case) basically reduced 15-ish lines of code down to a single line of code I only barely understand and could not have written myself. So that kind of success, to me, is meaningful.

Second, and tied to that, I had to test that code to make sure it worked properly. And there is a small glimpse in this of what often does not happen with AI, meaning I had to know enough about this app, and the code, to ensure that what the AI made was correct. In this case, it was. But in another similar episode, AI (I can’t remember which I used, but I guess it doesn’t matter) created similarly more concise code than what I had done, but in testing that code, nothing worked. Finally, I stepped through it with the debugger and discovered that the AI-created code was completely wrong. I had trusted it, stopped thinking about it, and then had to go back and fix it.

I guess there’s a third finding in there, though these are getting vaguer as I go: I feel like I could easily fall into this trap almost anywhere with AI, and that it is the trust that develops over time that allows that to happen. Many people evaluated AI early on, found it lacking, and then never looked again. But many more finally went down this path, were successful, and then stopped thinking critically about what AI was doing for them. Me falling into that so easily is a bit alarming to me, as I pay attention to this stuff. Most mainstream users won’t.

Anyway. Overall, my experience using AI for coding has been extremely positive. But coding is a great use case for AI because each language and framework is a finite set of data, and AI is good at grounding itself in that kind of thing. This gets more problematic in other, less concretely defined areas of life. As in…

On TWiT at the weekend, there was a discussion about AI being used in psychiatry and psychoanalysis. One of the guests is a qualified psychologist (Georgia Dow) and she said it is a great idea, but we are a long way from being able to use AI in that realm for serious problems. She said, for simple things, like arachnophobia, it could be useful, likewise talking to an AI psychoanalyst in the middle of the night, if you suffer from insomnia, could be useful, but AI just isn’t mature enough for anything that deals with the “dark places” of the mind, because of hallucinations.

You don’t want it to hallucinate and give the person looking to end their life an answer like jump off a building…

Right. But with the understanding that there will always be unethical people doing the wrong thing–like the infamous story about the lawyer who submitted an AI-created filing to court that cited hallucinated case precedents–most therapists (or psychiatrists or doctors, or whatever) are not going to use AI that way. My doctor references web-based information during my visits now, and that will shift to AI in time, of course. But it’s often for fact checking or looking up something specific. The people who come into these offices now are more informed than ever thanks to Google Search and whatever else, and we expect this from our doctors now too.

I guess the fear here is that many will simply self-diagnose using AI, especially since many seem to develop oddly personal relationships with these things, something that felt like a phenomenon only for older people with Alexa, especially, but probably other previous generation digital personal assistants, too. We’re all heard horror stories about an AI telling someone that they should kill themselves, or just be incredibly negative, which can be devastating for someone with serious issues. But as with anything technology-related, it improves, and in this case very quickly. I think it will eventually be made safe, if only because that will become a selling point for specific AI models and chatbots.

That has been my argument all along for any business use-case. There is a lot of potential in LLMs, but as long as they regularly hallucinate and their answers have to be double checked every time, they just are ready to be let lose on business processes – what if it decides that it needs to add a bit of sulphuric acid to the batch of shampoo it is currently producing? Or the brake master cylinder for a car would be cheaper to produce if it left out the rubber seals?

This goes back to my coding example above. In my case, I wrote code and asked AI to make it more concise and/or efficient. But for other tasks, many will start with AI and then edit or check the output themselves. I’m not sure how regularly AIs hallucinate, but the hit rate or whatever is only problematic in certain scenarios. If I ask an AI about a fact and it gives me the wrong answer, that is a problem. If I ask AI to create an image of whatever description and the thing it comes up with is humorously wrong, that might be considered creative. And it could serve as the basis for further work that might not have happened otherwise.

Whether you’re starting with AI or fine-tuning something a human created with AI, I see major efficiencies there. We have a friend who is a graphic artist, and I’ve referenced her regarding the AI-created images I use on the site, and how she (like any human) would be unable to fulfill that job. But she’s also someone who likes to post to social media, and she’s creative, but she’s not a natural writer. And so she’s started using AI to create text to accompany her posts. And that’s a great example of using AI for something that you may find difficult or uninteresting. Whether she checks that text before posting is kind of up to her (I don’t know, in this case). But the one thing she can rely on is that this will only get better, and it’s been getting better so fast that whatever worries we have today may be obviated very quickly. One could argue that she started using AI in this way at least in part because it did get good enough for this work.

I think that researchers should be looking at the error problem as the highest priority, producing bigger and “better” models that still have this basic flaw is missing the point, yes, we can communicate more easily, they can provide more complex answers, but if those answers are still full of errors, they still aren’t of much use for serious work.

I think you’re underestimating how much AI is already being used for serious work. This will vary by job/industry and whatever else. But it’s happening all around us. But I do agree that AI models need to improve in a safeguarding sense, as noted above. And that while most AI model makers talk about this now, it will become all the more important as these models are more similar and commoditized.

But I’m at least self-aware enough to recognize that I’m being somewhat hypocritical here as I don’t use AI for my day-to-day work at all. I’ve even started days thinking, OK, this time, I will do … whatever with AI. Summarize a corporate earnings report or a YouTube video transcript, something. Anything. And then I never do that.

? Yes, it matters

jrzoomer asks:

Paul with all the recent antitrust talk it got me thinking. Back in the 90s Microsoft beat Netscape, and made this such a high priority that some actions they took to ensure this were anticompetitive, to their detriment. I’d like to know what your thoughts are looking back at the browser wars with the lens of where we are today. I could imagine an alternative future if Microsoft just let things play out more competitively (and not invited the DOJ to investigate)!

Yes. I think about that fairly regularly, actually, and I’ve been thinking about it a lot lately, in part because I’ve been (re)reading various Microsoft history books and am looking at updating Windows Everywhere this year. In the series that became that book, I came to the conclusion that Microsoft lost the script in the wake of its Internet Tidal wave turnaround, about two years later, I guess, when it cancelled plans to fully embrace web technologies and went back to proprietary technologies. The most obvious result of this work was .NET, which is terrific now. But it was created to “embrace, extend, and extinguish” web standards and Java, technologies that had been created elsewhere and threatened Windows. Microsoft’s response to technologies that it feared would make it irrelevant was to just make itself irrelevant instead.

In the end, the open web, and web apps, were what mattered, not really so much the browser. All these browsers became commoditized, you can use whichever you like, and though IE did win back then, even today Microsoft doesn’t have market leadership in the web browser.

The battle between Netscape and Microsoft for the web browser is classic on so many levels. But as much of history, the real story is more nuanced than many remember. For example, while Microsoft did well-worn strategies to hobble Netscape and slow its progress, Netscape also made strategic errors of its own, and those did contribute to its loss. Also, Internet Explorer wasn’t all that interesting at first, beyond the weirdness of Microsoft making a browser at the time, but by IE 3.0, it had surpassed Netscape functionally, and that gap widened over the next few years. By the time IE beat Netscape, it was the better product.

When Microsoft was accused of antitrust violations by the DOJ in the late 1990s, it employed what is now a familiar argument with Big Tech: Though it had no real competitors, and a rich history of killing any competition that had been foolhardy enough to take it on, Microsoft argued otherwise. It said that technology moves fast, and that there was always some new company, technology, product, or service or whatever that would obsolete its dominant product overnight, and that regulation was thus unnecessary.

In Microsoft’s case, it invested $150 million in Apple specifically to keep that company afloat at a time when it was necessary to demonstrate there was competition. But in its court filings, it also cited non-events like BeOS and Linux on the desktop as threats, alongside Java, web browsers, and whatever else.

On some level, Microsoft was right. You can look at Google, a company that just made $90 billion in revenues in one quarter, and see both its dominance and the inherent threat in OpenAI and other AI makers to that business. But the possibility that some competitor might obsolete a dominant company/product doesn’t obviate or excuse the illegal business practices that Google or Microsoft employed, in both cases over a decade or more. We can’t forgive their sins just because they may some day be defeated.

And that’s the problem, Microsoft was wrong, too. Microsoft had so successfully killed off the market leaders in every sector in technology, from Ashton Tate to WordPerfect to Novell to Lotus to IBM and everything in between that it’s fearmongering claims about tiny Netscape ring as hollow today as they did then. This is a problem for Google and the rest of Big Tech today, as well. When you can leverage your dominance so thoroughly that most potential competition never even occur, you can hide the crime. Netscape was interesting because they tried, got the standard Microsoft pushback, and then complained to the government. We got to hear about something happening in real time instead of later, when it was too late.

So Paul what do you think, did it really matter that Microsoft beat Netscape?

Yes, very much so.

We can’t possibly know what the world would be like today if Microsoft had tried to behave ethically in the late 1990s, or if it had stuck with its embrace of open, web-based technologies. But we do know what happened because it did not. And one of the biggest changes to come out of this era is that Microsoft formalized the notion that operating systems would have web browsers built-in. The company argued that this was similar to adding networking capabilities, but that’s nonsense. Web browsers are applications, but they’re also an entire stack, from the rendering engine to the UI to the feature set, and they compete not just with other apps, but with app platforms. They are, in essence, OSes. The reason Microsoft–and then Apple and Google–built web browsers into their platforms was to prevent them from obviating their platforms.

Today, Apple, Google, and Microsoft all use their built-in web browsers to steer users in whatever direction is best for them, the platform makers. Apple prevents rival browsers from using their own rendering engines, not because of the stated reasons, but because they don’t want the competition from faster moving companies that might create a better product and lessen users’ reliance on the App Store and its lucrative fee-based revenues. Microsoft runs Edge when you perform certain tasks in Windows 11, even when you’ve configured a different default browser. These are anticompetitive acts, and unjustified from a technical perspective.

If Microsoft hadn’t created a web browser, this would never have happened. That doesn’t mean the world would be “better” or whatever. But you can see how Microsoft was both right and wrong about the future. Yes, competition did come along that lessened the dominance of Windows, in this case from the web and from mobile. But that took many years, and you can effectively argue it only happened because of regulation. Had Microsoft not been distracted by its back-to-back antitrust cases, the more heterogeneous world we have today, with Amazon, Apple, Google, and Meta all vying for our dollars, might never have happened.

Microsoft needed to be stopped, and it was. Today, Big Tech needs to be stopped, and it’s finally happening. And yeah. It really matters.

? Inevitable

train_wreck asks:

The Microsoft/Open AI partnership seems outwardly structured to avoid regulatory action. Do you think this will hold? If Open AI were to at some point end up with Google Chrome for example, could that trigger more scrutiny against the partnership?

One of the big debates in antitrust in the modern era is that these laws somehow can’t keep up with the pace of technology. But each time a Big Tech company is accused of antitrust violations, we learn the same lesson: Those 100 year old laws are applicable to technology and they work. The business practices we see with Big Tech parallel the anticompetitive behavior we saw from dominant companies in the past, no matter the industry. The center holds.

But Microsoft clearly orchestrated its partnership with OpenAI to avoid scrutiny, and this is an intriguing case of existing laws not working. It has nothing to do with technology, of course. But it’s interesting that Big Tech, not just Microsoft, is rethinking how it does things in this temporary era of heightened scrutiny.

Last month, the UK CMA cleared Microsoft’s partnership with OpenAI after a 16-month investigation. The reason it did so is interesting: Microsoft and OpenAI had just announced a surprising change in their partnership that allowed the latter company to seek outside infrastructure partnerships but with a Microsoft right of first refusal. We have to speculate why they made this change, but the leading contender is that the cost of building out its AI infrastructure was finally catching up to Microsoft, and because of the way their partnership is structured, it will still receive revenues when OpenAI workloads run elsewhere. But this, somehow, was enough for the CMA: In its view, this indicated that Microsoft did not “control” OpenAI behind the scenes. The partnership was not a secret merger situation. But the laws in the UK and elsewhere weren’t designed to handle this type of relationship, which is in a weird gray area between an outright merger/acquisition and a standard business relationship.

This is happening all over Big Tech now, with each of the major players having some stake in one or more of the upcoming AI players, all designed to side-step regulatory oversight. But regarding Microsoft and OpenAI specifically, it always felt weird how much Microsoft was relying on that company, and when the Sam Altman drama of late 2023 unfolded, the danger became clear. Since then, Microsoft has raced to replace OpenAI at every level possible, all while using as much OpenAI technology as possible. But there are two things that really stand out to me here.

One, these companies are partners, but they’re also competitors, and it’s interesting that Microsoft will undercut its partner by doing things like giving away capabilities that OpenAI charges for. This is reminiscent of its strategy against Netscape, per the above conversation. Netscape charged companies for its browser, but Microsoft gave it away. That’s another thing that came out of Microsoft beating Netscape. Now no company can charge for a browser, and that’s part of the reason the platform makers that also make browsers all have huge advertising businesses.

Two, the Microsoft and OpenAI relationship is very much like the relationship that Apple has with any partner. While it needs that partner, it does what it has to do, but you also know that it will drop that partner like a bad habit as soon as it can. Companies like OpenAI go to Microsoft (or Apple) knowing this is the inevitable future, but the goal is to be as necessary as possible for as long as possible so that they can grow. And the possible outcomes are many. Before our current regulatory era, these companies might be acquired by the bigger company, or by one of its competitors. Or they become so valuable that they go public, and you can see the first step in that direction now in OpenAI’s attempts to restructure the company as a for-profit firm.

Whatever happens, this one is going to get messy. And it’s going to end badly.

? Awkward

red.radar asks:

Co-pilot seems to be integrated as a feature into every MS application and even a stand-alone experience on all platforms. There are buttons everywhere especially in high traffic areas of mobile UIs. At least the co-pilot button in MS-Paint brings up image creation capabilities which seems relevant. But in outlook it just seems like a basic chat/query interface for general info. It doesn’t seem tuned to bring in an experience that is tailored for the use case of the application. It feels … awkward.

Yeah. With Copilot, especially, there seems to be a “damn the torpedoes”/”throw it on the wall and see if it sticks” mentality as Microsoft races to figure out how to best integrate this functionality into existing products. But that’s sort of the point, too, to give them a bit of credit. The name Copilot literally makes sense because the first generations of these AI implementations have to work with existing apps, many of which were designed decades ago. And so shoehorning AI into them isn’t always elegant or obvious. (This is the “besides” application structure that Stevie Bathiche introduced us to in 2023.)

It makes me wonder…the behavior feels desperate and almost un-planned. Like an org is desperately trying to find a way to make it stick… rather than tailer the exerience to the use case of the applications they want to integrate AI into.

In time, this will evolve into more integrated experiences in time, and new apps and services that are designed with AI from the get-go. But yes, for now, it’s a matter of trying things with existing apps and seeing how it goes. There are lots of examples here. Microsoft promoted Recall as the marquee feature of Copilot+ PCs in mid-2024, but today I would argue that Click to Do and Copilot Vision, both of which integrate with whatever is happening onscreen, are more useful for more people.

But there are also little examples of specific AI functionality integrated into existing apps that work well. For example, Snipping Tool at some point picked up OCR capabilities via a “Text actions” button that would appear in the app after you took a screen capture. (This lets you scrape the text out of a shot and then copy it to the Clipboard.) But in a coming update to that app, the “Text actions” button also appears in the initial app view, the capture bar, at least on Copilot+ PCs. This now integrates with Click to Do so that you can arbitrarily select a region onscreen in real time, no capture required, and then get at that text. So that’s a simple evolution of a good/useful AI feature, and it came together relatively quickly. There are less successful examples that speak to the chaos of this approach, of course. The Copilot app in Windows is perhaps the poster child there.

All that said…How is Co-pilot doing compared to the other AI models? I feel like Co-Pilot is on the path to being the Bing of AI models as Chat-GPT continues to be the front-running leader… Am I letting my frustration cloud my objectivity?

No, that future is a very real possibility. Two years ago (!), I argued that when everything is AI, nothing is AI, meaning that as Big Tech added AI everywhere, it would just become table stakes (I hate that term), an expected set of functionality that all platforms, apps, or services would offer. And that it would thus do nothing to level the playing field or upset the market in any way. In Microsoft’s case, this means that Microsoft 365 would still be dominant with companies, especially in the Fortune 500. And that Windows would continue to have whatever market it’s had in recent years.

But these companies don’t want the status quo to continue, they want to expand their reach. And I agree that Microsoft has a problem. Outside of Microsoft 365, especially, its AI is not of particular interest. Few are opening web browsers and going to the Copilot website, let alone Bing. This isn’t conjecture: I’m working on a story about Chrome and what might happen to the browser in the wake of Google’s search-based antitrust loss, and there is some very interesting data coming out of the remedy hearings.

Google provided this slide during the hearings this past week to show that Gemini was part of a very competitive market for AI chatbots, with OpenAI the dominant player.

But do you notice which AI is missing? Right, Copilot. Now, that might be a purposeful slight given the historic relationship between Google and Microsoft. But I think Copilot usage outside of Microsoft’s platforms–Windows, Microsoft 365, and maybe Edge–is tiny to nonexistent. And that when people do engage with an AI model on the web or mobile, it’s typically ChatGPT. Gemini, Meta AI, and Grok are all platform-based usage, like Copilot. But ChatGPT is the Chrome of AI, meaning that it is super successful regardless of platform makers building their own AIs. (Google was overtly trying to show some growth with Gemini here, too, but most of that must be all the free Gemini giveaways across Android, Pixel, and Chromebook/Chromebook Plus).

(UPDATE: Here’s some more proof of low Copilot usage: 20 million average weekly users for Copilot vs. 400 million for ChatGPT. –Paul)

In the Microsoft space, we recognize that Copilot is the rare example of a good brand from the software giant, but we have to also remember how little this company matters to consumers. If Copilot’s fate is that it’s something we use in Windows and with Office, then it’s just work, and that’s fine, it will have some level of success. But consumers–people–are using ChatGPT, just like they’re using Chrome. And I don’t see that changing, certainly not in Microsoft’s favor. The good news is that Microsoft does benefit from ChatGPT’s success. And that in losing, Microsoft could still win by hosting that AI infrastructure for OpenAI and others that do create popular end-user services.

? Know thyself

facelessghost asks:

Your Also-Ran article struck a chord with me (and, judging from the comments, several other readers). Not just in tech, it seems I’ve spent much of my life supporting and pursuing the also-ran option.

Yeah, we all have. This is the curse of being in the Microsoft ecosystem, and tied to some of antitrust commentary above, the shift to a more heterogeneous personal computing market means we went through a long period of Microsoft trying and failing with more products than I can now remember. I think I owned all of them, though, and I felt pretty strongly about more than a few of them.

Right now I’m in the market for a new laptop, and it’s come down to a MacBook Air or a Surface Laptop, and I can’t seem to bring myself to go with the former. Part of that is my dislike of Apple,* and part is my deep familiarity with Windows, which I use all day, ever day at work (but I am not an Enterprise, right?).

In any event, my question is this–in the consumer/personal market, is Windows in danger of becoming the also-ran option? I can’t imagine Windows going away in any foreseeable future, but I can imagine a world in which Windows is something we all use at work, but at home–to the extent we use computers at all, rather than phones or tablets–it’s nothing but Chromebooks or Macs. Is Windows headed for a fall? Am I once again sticking with the also ran option?

In the PC space, Windows still controls about 72 percent of the market by usage share, if you believe StatCounter’s numbers, compared to roughly 15 percent for the Mac and, less believably, almost 4 percent for Linux. (And less than 2 percent for ChromeOS.) But that figure is closer to 90 percent if you look at unit sales, with the Mac landing at less than 9 percent of the market. So Windows is not an also-ran in the PC market, which is what you’re concerned with here. (It is the smallest of the three big personal technology ecosystems, with Apple and Google more dominant as we factor in mobile devices and web.)

When it comes to a purchase like this, you’re not just looking at the hardware and OS. It’s really about how you use the computer, and why, and ecosystem can be a major factor. One of the big advantages with Apple is that the Mac integrates so well with its other devices, especially the iPhone, and so if you’re an iPhone user that could be a draw. On the Windows side, Phone Link has evolved into something almost indispensable if you’re an Android user, and I rely on that now for text and calling through the PC, plus some notifications and photo sharing functionality.

As Dirty Harry once observed, “A man’s got to know his limitations,” which to me is really just above self awareness. And in this case, that means you knowing how you’ll use this machine and which makes more sense in your use cases. I have struggled with the Mac since Apple announced Mac OS X and started shipping beautiful laptops starting with that Titanium PowerBook. I have always loved Apple’s hardware, and I’ve always been intrigued by what Apple’s done with the OS, but I have also always gone right back to Windows. I find it more familiar, more logical, and more efficient. I experienced this most recently, and maybe most viscerally, with the MacBook Air M3 I bought last year. There is nothing to not love about the hardware, and the performance and battery life is incredible. But I very much prefer my Surface Laptop 7. I’m using it to write this, in fact.

I don’t know where you land on the other factors in this matrix of decisions that will result in you getting whichever computer. I do know that you will love either laptop, mostly. And that both are great choices. But I wouldn’t worry about Windows being an also-ran. In the PC space, most active development is occurring with web apps, and those work fine everywhere. Perhaps your dislike of Apple puts this one over the top.

☀️ Oaxaca

AnOldAmigaUser asks:

How was Oaxaca? Any recommendations? My wife is very interested in visiting on our next trip to Mexico.

I’m going to write something about Oaxaca soon, but the short version is that I fell in love immediately, and we will be going back. Oaxaca was something I had not experienced in the other cities we’ve visited here–Guanajuato, San Miguel de Allende, Puerta Vallarta, and Puebla–and a reminder of how we used to travel before we settled on Mexico City. Oaxaca is a quick flight from Mexico City, about 40 minutes in the air, so that part of the trip is easy enough. I seriously doubt we would have bought this apartment had we visited Oaxaca first.

More on that soon, but I do think it’s important to point out that no place is perfect. Oaxaca is extremely hot, it was in the high 80s and low 90s when we were there, and the average high temperature at this time of year is above 90 degrees. But the interesting thing about attraction, whether you’re talking people, places, or anything else, is that the things we love have flaws we can overlook. Paris is my favorite place on earth, and Washington D.C. is my favorite city in the U.S., but both have all kinds of problems, and one they share is extreme weather: Both are incredibly cold and dreary in the winter and unbearably hot and humid in the summer. I love them both.

Oaxaca has other issues. The airport is a glorified bus station. There are no Ubers, as is the case elsewhere in Mexico, and when we tried Uber when we arrived, there were no cars available. (We’d never seen that before.) When Uber works, you get a taxi, and our few experiences with that were negative. There is a lot of traffic. And there is a horrific amount of graffiti everywhere, from beautiful historic buildings to churches and everything in-between. I don’t mean street art, I mean horrible, pointless, spray-painting. Everywhere. I can’t believe the city allows this, or puts up with it, and I can’t believe anyone would do such a thing. It’s not like that here in Mexico City, or anywhere else we’ve been in this country.

But … I love it. The food, the architecture, the people, all incredible. I can’t wait to go back.

?️ Chrome’s future is so bright, I gotta wear shades

AnOldAmigaUser asks:

For all the talk of Google having to divest Chrome, does it make sense for any company to buy it? The value of Chrome for Google is the amount of data it collects on the people that use it. Clearly, all of the backend bits that hoover up browsing information will have to be, at a minimum, pointed elsewhere. Otherwise, it needs to be removed. This would make “Chrome” simply a brand to other companies, unless their plan is to become the next Google which would defeat the whole purpose of the ruling.

This is a rare double whammy, so apologies for this, but as noted above, I’m writing separately about this topic as well. But let me provide some high level points to a topic that is still ongoing and evolving.

When Google announced Chrome, it explained that it’s entire business was based on the web, and that it was thus beholden to third parties, primarily Mozilla, Microsoft, and Apple at that time, that had different priorities and perhaps couldn’t be relied on to deliver the right capabilities, especially performance, that it needed. So Chrome made sense, and it was immediately popular with users, and a wake-up call of sorts to other browser makers.

Today, Chrome is a vector for tracking and advertising, both core to Google’s business. It’s not clear how much of the $51 billion in “Google Search & other” revenues in the most recent quarter are directly attributable to Chrome, but it has to be a big percentage. (Bundling on Apple, Samsung, and other Android devices are likely bigger.) This is why DuckDuckGo, OpenAI, Perplexity, and Yahoo have all told the court this past week that they would buy Chrome if they could (or could afford to, in DuckDuckGo’s case). Chrome is so popular that they could replace Google Search with their own or a partner solution, and getting just a fraction of whatever revenues Google makes would be huge for those companies and others.

Chrome is quite valuable. The estimates vary, but a DuckDuckGo executive testified that his back of the napkin math valued the browser at about $50 billion. Bloomberg estimated Chrome’s value at $20 billion.

It seems to me that it would be easier to simply ban Google from distributing a browser, either in perpetuity or some defined timeframe. The Chromium project would then have to take over APIs and the extensions store…is that feasible? What happens with Chromebooks? Sans a browser, they are fairly pointless. Would this usher in the year of desktop Linux, since any distro with a browser is a Chromebook equivalent with a little more flexibility, and problematic system management?

Even if Google isn’t forced to divest itself of Chrome, Chromium needs to be controlled by an industry standards body, not a commercial company. And Google could simply keep using Chromium for its browser, if it keeps Chrome or not. ChromeOS is based on a ChromiumOS fork, for example. That could simply continue, with Google just being a licensee like all the other browser makers.

And even in a worst-case scenario, most users would likely continue choosing Google Search, and that means that Google will continue earning money from that business.

This is all certainly above my paygrade to decide, but I wonder if those making the decision are going to look at things holistically, or just go after Google with a chainsaw and no thought. This could be a watershed moment in reining in the surveillance economy a bit. I wish I had more faith that whatever is done is done solely to benefit the public, not merely to punish one company and replace it with another.

Punishing Google does benefit the public. It restores competition to a market that’s been artificially skewed for too long. Antitrust is about fair competition, primarily, and the biggest abusers, like Google, hurt everyone involved, from users to developers, partners, and competitors.

Then, there is Android…

Yeah. That feels a lot less likely to me, and even in the DOJ’s recommendations, it feels like a “what if” scenario should it’s proposed remedies not fully solve the problem. That’s smart: Forcing a divestiture is a big move, and forcing two like this is possibly unprecedented. By proceeding slowly, the government and the courts can give Google a chance to change its ways and if it continues the abuse, they can simply say the company did it to itself.

Anyway, I don’t see Google losing Android. But I do see them being banned from entering into any bundling agreements. I’m still mixed on Chrome. That is extreme, on one level, but it’s possibly necessary too. I will keep processing this, as there’s been new information every day this week. And while Google could make some ground up on appeal, it could also settle, with the idea of helping control its destiny. That’s the path I recommend. Letting go of Chromium would be a good opening concession.