Google Bard has had a fuzzy sense of my Washington Post career

Rejection hurts, but does it hurt more or less when a faceless conversational artificial-intelligence chatbot has done the rejection? And when the rejection comes in the form of erasing a large chunk of one’s career?

Trying out Google Bard, the “large language model” AI that Google opened up to U.S. users in March, forced me to think about questions like that. It started when I saw a Mastodon post from my Fast Company editor Harry McCracken recounting how Bard reported having no information about him–and then, after Harry asked the exceedingly resume-specific query “Who’s the guy who worked at PC World and then started Technologizer and then worked at TIME and now works at Fast Company?,” Bard responded by giving credit not to him but to my old Yahoo Tech colleague David Pogue.

Screenshot of Google Bard attempting to answer "Who is Rob Pegoraro?" and "Who wrote a personal-tech column for the Washington Post from 1999 to 2011?"

I had to repeat the experiment, but I didn’t see quite the same results. While I got the same “I do not have enough information about that person” brushoff to a who-is query about myself, asking “Who wrote a personal-technology column for the Washington Post from 1999 until 2011?” yielded a flattering response that said I currently write for USA Today (not any more) and Wired (never have); had earlier written for the New York Times, the Wall Street Journal, and the Atlantic (the last is true); and had won a National Press Club Award for Consumer Journalism and the Gerald Lob Award for Commentary (if only).

A few days later, I tried again and got a different set of results. Once again, Bard couldn’t say who I was–but this time, asking who wrote a personal-tech column for the Post until 2011 led to Bard naming former WSJ tech columnist and all-around mensch Walt Mossberg. Walt has won a Loeb award, so maybe Bard’s earlier “hallucination” (the current term of art for a conversational AI making shit up) involved a fusion of his c.v. and mine?

A few days later, I took myself out of this equation by asking a Google rep at a tech-policy event in D.C. Monday to repeat my searches. She, too, saw Bard decline to answer a who-is question about me and then reassign my Post column to somebody else–this time, Wired’s Steven Levy, another member of the graybeard tech-writer contingent. The Google rep clicked the thumbs-down “Bad response” button (which I thought I’d done earlier, but maybe not) and invited me to write the correction.

But my pointing out that I wrote the Post’s personal-tech column over those years apparently needed more than three days to get processed. Because when I tried the same series of questions Thursday, I got the same “I do not have enough information” answer to “Who is Rob Pegoraro” and a comparable erasure of my Post history:

Steven Levy wrote the Washington Post’s personal-tech column from 1999 to 2011. He is an American journalist and author who has written extensively about technology and the internet. He is the author of several books, including Hackers: Heroes of the Computer Revolution, Insanely Great: The Life and Times of Macintosh, and In the Plex: How Google Thinks, Works, and Shapes Our Lives.

Once again, I clicked the “Bad response” button but this time also talked back to Bard: “Try again.”

The answer this time:

Rob Pegoraro wrote the Washington Post’s personal-tech column from 1999 to 2011.

Victory at last! Or maybe just for now.

Advertisement

Late or never Android updates remain a problem

Here’s yet another unintentional benefit of my shattering my Pixel 5a’s screen last weekend: an opportunity to reacquaint myself with how slowly many Android smartphone manufacturers still ooze out Google’s system updates.

This is not a new problem, as I can see from re-reading a piece I wrote almost 10 years ago that’s aged a little too well. I had thought that architectural changes Google made to Android starting back in 2017 would have put a dent into this problem by removing much of the recoding work from manufacturers. But dusting off the budget-priced Android phones I reviewed for CNN Underscored early this year (most of which I had not yet returned to the companies responsible, because my desk is a mess) revealed the error of that thought.

Photo shows Android phones stacked on a wooden floor, each showing their software-information screen. The Samsung Galaxy A13's screen is most visible, showing it's running Android 12 with the July 1 security patch.

After multiple cycles of checking for updates on these six phones, installing these updates, rebooting these phones, and checking for updates again until every device reported it was current, here’s where they wound up:

  • Moto G Power: Android 11, August 1 security update
  • Nokia X100: Android 11, August 1 security update
  • OnePlus Nord N200 5G: Android 12, September 5 security update
  • Samsung Galaxy A13 5G: Android 12, July 1security update
  • TCL 20 SE: Android 11, August 1 security update
  • TCL 20 Pro 5G: Android 11, April security update

The current month is October and the current Android version is 13, so the problem should be immediately obvious. And not only did none of these devices have the Android release that I installed on my beloved, now battered Pixel 5a in the middle of August, only one of these devices had Google’s latest security fixes–and only two had the Android release that Google shipped a year ago.

The good news, such as it may be, is that a low price doesn’t condem an Android phone to obsolescence. The A13 sells for $250 and the N200 $240, but both have aged better, software-wise, than the pricier Android devices in that review. You may want to consider that a factor in favor of OnePlus and Samsung if you’re shopping for a low-cost Android phone–while the lagging performance of those other vendors should rate as a serious strike against them.

The multitasking interface in iPadOS 15 is not aging well for me

It didn’t take too long after I installed iPadOS on my iPad mini 5 for me to restore order to my app-icon grid–even if I’m still tweaking that arrangement and dreading the moment when the next iPad system update sends it higgledy piggledy. But another part of Apple’s tablet operating system continues to grind my gears: its multitasking options.

I can’t fault Apple for trying to make this UI more discoverable. In the previous release, I had to look up how to run one app on a third of the screen and leave the other two-thirds to another app every time I wanted to have the clock app and my notes visible side by side for a virtual panel. But in iPadOS 15, I have the opposite problem–the system keeps thinking I’m trying to split the screen between two apps when I have no such intention.

The most common scenario involves me wanting to go to a different site in Safari, when tapping the browser’s address bar routinely invokes the three-dot multitasking button that Apple added to iPadOS 15. That bit of chrome may stay out of the way more often on a larger-screen iPad, but on the 7.9-in. display of my iPad mini, it’s a different story. There, only a few millimeters of screen real estate–either from the top of the screen to the address bar, or between the center of the address bar and address-bar controls like the text-size/display/privacy button and 1Password’s button–seem to separate me from successfully entering a Web address or having the multitasking button thwart that attempt.

The other involves a situation almost as common: iPadOS flashes a notification, and I swipe down to see what it was. From the home screen, this continues to work as it did before–but in an app, iPadOS keeps acting as if I’d meant to invoke the Split View multitasking display by tapping that dreaded three-dot button. Eventually, I will reprogram my muscle memory to swipe slightly off-center to avoid running my finger across that ellipsis icon, except the home-screen behavior keeps telling me I don’t have to change.

So here I am, more than six months after installing this update, and I’m still thumb-wrestling my way around one of its core features. And I’m not alone in feeling this irritated, to judge from my mom’s review of this wayward user experience: “the most distracting thing in the world.” She’s right, and Apple’s wrong.

A long wait for an app notification

Twenty-one months ago, I installed the Virginia Department of Health’s COVIDWISE app on my smartphone and urged everybody reading that post in Virginia to go and do likewise. Back in August of 2020, I expected that this app developed with the Apple-Google COVID-19 exposure notifications framework would soon be warning me that I’d been near somebody else who had tested positive and had then used this app or another built on that foundation to send a thoroughly anonymized warning.

But the notifications of possible exposures didn’t appear, even as the U.S. suffered repeated waves of novel-coronavirus variants and the positive-test rate in Northern Virginia shot up above 30 percent at the start of this year. And as I got my first vaccination, second vaccination and booster shot, the continued silence of this app bothered me less and less–to the point that I briefly forgot to activate it after moving from my Pixel 3a to my Pixel 5a.

That silence ended Thursday morning, when my smartphone greeted me with a notification of a probable exposure. “You have likely been exposed to someone who has tested positive for COVID-19,” the app told me. “COVIDWISE estimates that you were last exposed 5 days ago.”

The app further informed me that “Most people who are fully vaccinated and free of COVID-like symptoms do not need to quarantine or be tested after an exposure.” Fortunately, I had already self-tested negative on an antigen at-home kit Wednesday morning to verify my health before heading to the Hack the Capitol security conference.

Because this app and others built on the Apple/Google code don’t store location data, I can only wonder when this possible exposure happened. And since five days ago was Saturday, when I flew home from Latvia via Munich and then Boston, I’m looking at thousands of miles of possibility. A second notification from COVIDWISE referencing North Carolina’s SlowCOVIDNC app suggests that my possible exposure source lives there, but the privacy-preserving design of this system ensures I’ll never know for sure.

A five-day turnaround, however, now seems quick after seeing three people reply to my tweet about this notification to report that they didn’t get their own heads-up from one of these exposure-notification apps until 10 days after the possible exposure–a uselessly long lag. My conclusion from those data points: Get vaccinated and boosted, because that will do more than anything else you could possibly undertake to ensure that receiving one of these exposure alerts remains a drama-free experience.

Flickr’s Android app still needs some work

Today brought yet another rerun of a mobile-app routine that I’d like to see closed out: I saw an update for Flickr’s Android app waiting on my Pixel 5a, I installed the update, I went out and took some photos, and I saw that Flickr still isn’t saving location data when it backs up pictures automatically.

As bugs go, this one is not too consequential. But it’s also gone unfixed since late December, when I first noticed it. Flickr’s customer support promptly responded to that tweet, asking me to send in a sample photo, and the response I got hours later let me hope for a fairly quick resolution.

“We did some testing with the image you provided, and got mixed results upon uploading to a private test account that we use for cases like this,” the rep wrote. “The issue seems specific to the app, and possibly to the new update of 4.16.6.”

But the Flickr app has since seen multiple updates–it’s now up to version 4.16.15–and I’m still seeing automatically uploaded photos stripped of their geotags. That seriously erodes Flickr’s use as an unlimited-storage image backup vault; fortunately, I mainly employ Flickr for its original purpose of photo sharing, and for that I can use the share menu in Google Photos to post pictures, GPS data intact, to Flickr.

Flickr reps have remained a pleasure to deal with over e-mail, and the most recent one added three free months to my Pro subscription, which was a nice gesture.

But I also don’t feel that I can get too mad here, given that Flickr is my only major social-media platform that isn’t the property of a tech or media conglomerate, having been rescued from Yahoo’s erratic stewardship by the privately-held photo-sharing firm SmugMug in 2018.

And not only does Flickr have the advantage of being Not Google and Not Facebook, its support for albums has no equivalent on Instagram. And its support for Creative Commons licensing (my photos are free for non-commercial use while commercial users are welcome to pay) and groups of like-minded photographers (for instance, Capital Weather) have no equivalents at either Instagram or Google Photos.

I don’t even mind having this expense of a Flickr Pro subscription, now $71.99 a year and increasingly hard to avoid, in my Web-services budget. And while I would not turn down getting a few more months free, I would just as soon see Flickr fix this bug and restore this app to complete functionality.

RSS lives, but not equally well on all of my devices

Many of my daily digital habits have changed over the past decade and change, but one has not: I still check an RSS app to see what’s new at a couple of dozen sites. Even as the social-media and news landscapes have gone through multiple cycles of destruction and reinvention, the standard originally called Really Simple Syndication remains one of my preferred ways to know when a favorite site has published new posts.

The core advantages of RSS (which you may find easier to call “Web feeds” if you’re trying to cut back on tech abbrevations) remain unchanged from what I wrote here in 2015:

One is control: my RSS feed only shows the sites I’ve added, not somebody else’s idea of what I should know. Another is what I’ll call a tolerance of time: A site that only posts an update a week is less likely to get lost when it occupies its own folder in the defined space of my RSS feed.

The third, maybe most important feature: Nobody owns RSS.

That last angle allowed me to move on from Google’s capricious decision to kill its Google Reader RSS service–on a scale of that company’s faceplants, it’s up there with Google Glass–and switch to Feedly’s RSS platform.

The individual apps I use on my various devices, however, have changed a bit since then.

My favorite among them is the free and open-source NetNewsWire–the iPad version of Brent Simmons’ labor of love more so than the Mac edition. On my tablet, this app displays a simple count of unread articles (which, sadly but typically enough, is once again near triple digits as I write this) and lets me read each one in full without switching to Safari. On my desktop, NNW also displays that app-icon badge count but requires a switch to Safari to read a new post if the site, like most, doesn’t publish a full-text RSS feed–meaning I’m taken out of my reading workflow and then have yet another tab to remember to read and then close.

On my Pixel 5a phone, I use Feedly’s Android app. The free version stuffs some ads into feeds, which I don’t mind all that much since they’re so easy to skip, and it does let me read entire posts in its built-in browser instead of requiring a switch to Chrome. But Feedly’s app doesn’t even try to support Android’s notifications system (if only that OS feature supported not just adding a colored dot to an app icon to indicate new content but a count of unread items as iOS does), so it’s too easy to ignore its green icon on my home screen.

On Windows, I have a different problem: NextGen Reader, a fave of critics in the previous decade, still works fine–with the convenient feature of an unread-articles count on its taskbar button, but without in-app reading of original posts–but also appears completely abandoned. A decade after its debut in the Microsoft Store, it’s gone from that app market, its Twitter account has been silent since February of 2021, and the developers’ site now redirects to a generic parked-domain page.

And the other Windows RSS apps that I’d considered before settling on NextGen five years ago look equally moribund. So I don’t know what I’ll do if this apparently fossilized Windows app stops working–except, perhaps, reach for my iPad instead.

E-mail like it’s 2012: revisiting my Gmail filters

Several months ago, I spent too many hours hacking away at the 18 years’ worth of messages piled up in my Gmail account–because while I could live with paying for extra storage for Google’s backups of my own photos, I’d be damned if I was going to pay to warehouse random companies’ marketing pitches that were eating up far more of my free storage.

Screenshot of Gmail's filter, showing a menu of options that

That experience evidently wasn’t enough fun for me, because over the last couple of weeks I’ve dived into a corner of the Gmail interface I hadn’t spent any sustained time in since… 2012? Fortunately for me, Gmail’s filter interface hasn’t sustained any notable changes in at least that long, judging from how rarely it’s earned a mention in Google’s Gmail blog over the last decade.

Much as in 2012, this dialog lets me choose a message by factors like sender, recipient, subject, content, size and the presence of an attachments. Its next pane allows me to amplify the message’s importance by starring it or marking it as important, apply a label or file it under a category tab, or forward it to to an outside address registered with your Gmail account using an even more fossilized interface.

Those are the basics of e-mail management, and they did suffice to help me craft updated message rules that make my Gmail inbox less chaotic and keep my more consequential correspondence filed away neatly. Less e-entropy is a good outcome.

But the limits of the filtering user experience loom large among Gmail’s missing features. The one I keep coming back to is an equivalent to the “sweep” function in the Microsoft’s Outlook.com that automatically whisks matching messages into the trash 10 days after they arrive. But it’s also crazy that the entire filter UX is imprisoned in Gmail’s Web app–you can neither create nor edit nor view filters in Gmail’s Android and iOS apps, as if the last 10 years of mobile-first computing never reached whatever Googleplex building houses the Gmail developers.

Compare that thin gruel to the thoughtful mail-management tools surfacing in apps that actually have to win customers–I’m thinking here of a new Gmail front end called Shortwave that a team of former Google developers just shipped, but also of the Hey mail service. It’s not hard to think that Google could do a lot more with Gmail if it put serious effort into that work. It’s also not hard to think that Google must feel as comfortable in the e-mail market as Microsoft and Yahoo did before Gmail showed up in 2004.

Two belated desktop operating-system upgrades

Both Apple and Microsoft shipped major new releases of their desktop operating systems last year. And then I, somebody who writes about computers for a living, didn’t get around to installing either macOS Monterey or Windows 11 until this month.

I guess it looks a little embarrassing written out like that. But I had my reasons, the first among them being the absence of story assignments that would require me to familiarize myself with either OS. My regular clients all had other writers evaluating Monterey and Windows 11 and, as far as I could discern, they weren’t in the market for stories taking a closer look at any one feature in either new release.

A collage features Apple's round, stylized-mountains logo of Monterey and the square blue "11" logo Microsoft sometimes uses for Windows 11.

Further, desktop operating systems just don’t figure as much in my coverage, even factoring in the six years that elapsed between Windows 10 shipping (which I covered in July 2015 at Yahoo Tech) and the debut of its successor. My last piece focused on a macOS release seems to have been a June 2017 Yahoo post suggesting features for the upcoming macOS High Sierra (Apple, as is its habit, did not take my advice).

That freed me to take my own time, so I ignored my Mac mini’s suggestions to install Monterey after its late-October debut and instead waited to see if early adopters would report any glitches. Enough such reports emerged–in particular, with third-party USB hubs–that I decided I’d wait until the first update to Monterey shipped before I’d put it on my mini, on which I rely heavily on the USB hub built into the Philips monitor I bought with the computer.

My late-2017 HP laptop’s copy of Windows 10, meanwhile, didn’t even suggest installing Windows 11 when that update shipped in early October, in keeping with Microsoft’s phased rollout of this OS. But by the time Windows Update finally offered Win 11 sometime in December, it was too close to CES, and I didn’t feel like complicating my show prep by installing a new OS on a five-year-old machine.

After surviving CES, however, it was finally time. I installed Monterey on this Mac without incident a week ago, but Windows 11 threw one last obstacle at me: Both Windows Update and Microsoft’s PC Health Check app said my laptop was no longer compatible, citing an issue with its Trusted Platform Module security chip. And after re-enabling that in the HP’s BIOS, I had to wait another day for Windows Update to agree with PC Health Check that it was time for Windows 11–allowing me to install it Friday.

You may have noticed that I’ve now written more than 400 words in this post about these releases without discussing any one new feature in either. That is not by accident–and that’s why, now that I’ve finally caught up with 2021’s updates in 2022, I don’t feel like I missed out on too much by waiting to install Apple and Microsoft’s latest work.

CES 2022 travel-tech report: a new phone and a renewed laptop

Uncharacteristically light attendance at CES this week allowed me to pack uncharacteristically light. With so many tech-news sites canceling plans to send journalists to the Consumer Technology Association’s annual gathering, I knew I wouldn’t need my traditional CES accessory of a travel power strip to free up outlets in any crowded press room.

I also opted not to pack any of the WiFi hotspots I had sitting on my desk from the last update of Wirecutter’s guide to same. Even in the likely event of the show’s WiFi being its usual inadequate self, I figured I had sufficient backup bandwidth in the form of the new Pixel 5a in my pocket, the expanded mobile-hotspot quota on my account, and the T-Mobile 5G network my previous phone couldn’t use.

Photo shows my HP Spectre x360 laptop sitting on the wood floor of my home office, on top of which sit my CES badge, the laptop's charger, a USB-C cable to charge the phone, a USB-to-Ethernet adapter, headphones and my Pixel 5a phone.

My other smart move before heading out to Vegas was replacing my late-2017 HP laptop’s battery with an aftermarket unit, a bit of laptop surgery I did in October. All of this helped make CES much less of a gadget-abuse scenario in my return to covering it in-person after last year’s distanced, digital-only conference.

The Google Pixel 5a, the only new device in my messenger bag, acquitted itself especially well. On a good day, its battery can run well into the next afternoon, and even at CES–where I did rely frequently on the phone’s mobile-hotspot feature to get my laptop online–I never saw this Android device’s battery get into the under-33% state that would get me nervous. My charging the phone at lunch happened out of habit, not necessity.

I also quickly grew to appreciate how the 5a’s wide-angle lens helped capture some of the bigger exhibits on and off the show floor. The sole quibble I can think of: The phone reported that it restarted overnight Friday morning, and I’d like to know what caused that crash.

My HP Spectre x360, meanwhile, was one of the oldest items in my bag but felt much newer with that replacement battery. It was nice to sit down to watch a panel and not need a spot next to a power outlet. And for whatever reason, this computer ran much more reliably than it had at CES two years ago, without any mysterious reboots sometimes interrupted by boot-device-not-found errors.

Lower CES attendance, estimated by CTA Friday at “well over 40,000,” did not banish CES bandwidth struggles. My laptop did not always connect to conference WiFi networks–have I mentioned that Windows 10’s “Can’t connect to this network” is not a helpful error message?–but all three press rooms had abundant Ethernet cables. The $10 and change I spent on a USB-to-Ethernet adapter in 2012 has turned out to be an exceptional deal.

As before, I took all of my notes in Evernote, but this time the app generated a few note conflicts when I switched from phone to laptop and back. If I could click or tap a “sync now” button before each device switch, I would–but Evernote removed that bit of UI out of a belief that its automatic sync is now reliable enough to make it obsolete.

The other app I leaned on heavily during my time at CES was the conference’s own mobile app. I hadn’t bothered with that in previous years, but learning that CTA had hired Web Summit to provide this event’s digital platform made me want to try it. Unsurprisingly, the CES app looks and works like the Web Summit and Collision apps, so I didn’t have much to figure out.

As at those other conferences, I leaned on this app to manage my schedule while ignoring in-app connection requests in favor of the kind of networking impossible at last year’s CES: masked-face-to-masked-face conversations that ended with an exchange of business cards.

Android phone migration has gotten easier–except for Google Pay and Google Voice

Moving from my old Pixel 3a to my new Pixel 5a provided my smoothest Android phone-migration experience yet. I had much less home-screen housekeeping to do on my new device than two years ago, and one key Google app showed a particularly dramatic improvement. But then I had to deal with Google Pay and Google Voice.

Overall, Google’s instructions get across how easy process has become. Tap yes in the “Copy apps & data” button on the new phone, unlock the old phone, connect the two with a USB-C cable, tap yes in the old phone’s “Copy data to new phone?” dialog, then wait–about 21 minutes in my case.

A Pixel 5a showing the "Transfer accounts" screen in Google Authenticator sits atop a Pixel 3a showing the same screen in the same app.

Google’s Android-transfer system accurately reproduced my app-icon layout (the contrast with upgrading to iPadOS 15 did not escape my attention) and wallpaper, with the only missing item being a home-screen icon for Android Auto.

I did still have to wait for most individual apps to download off Google’s Play Store, and their new-phone user experience varied awkwardly. Some, such as Feedly, LinkedIn and FlightRadar24, didn’t need me to log back in; most demanded a new entry of usernames and passwords (made much easier by 1Password); a few required extra bouts of authentication.

One Google app pleasantly surprised me, given the sensitivity of its stored data. Google Authenticator previously required renewing each two-step verification code securing a site login as if your old phone had fallen into the ocean, an experience that Google security chief Stephan Somogyi in 2017 apologetically described to me as “a complete, total and unmitigated pain.”

But in 2021, an old phone’s Google Authenticator can generate a catchall QR Code for its saved accounts; scan it with the new phone’s copy of Authenticator, and you’ve got your one-time passcodes for those accounts ready there. Great!

And then two other Google apps showed how awkward this process can remain. Google Pay–not the mobile-payments app that debuted as Google Wallet, but the new release that shipped this spring and then required some non-trivial settings restoration–landed on the new Pixel 5a as if I had never used it before.

I had to start by typing in my cell number because this Google service relies on that for authentication instead of a Google account. As Ars Technica’s Ron Amadeo explained/warned back in March, this setup resulted from Google electing to build a new Google Pay off code optimized for the Indian market, where SMS authentication apparently reigns supreme. And then I then had to add back my saved credit cards, one at a time.

The last hiccup, I hope, came with Google Voice. The oft-neglected Internet-telephony app that I use for my work number seemed to be configured properly on the new phone, but then a journalist trying to reach me for a radio interview had her call go to voicemail. Eight times in a row. The answer turned out to be that Google Voice’s account settings had my number associated with two smartphones and two copies of the same number, a level of confusion that the system evidently resolved by not patching calls through to the newest device.

But now that’s squared away, and I think I can make it through the rest of this trying year without further mobile-app troubleshooting. I hope that’s the case for everybody reading this too.