home.social

#digitalservicenetwork — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #digitalservicenetwork, aggregated by home.social.

  1. October 14 – 20

    My “weeknotes” capture events, thoughts, and other items from the past week, mostly focused on work. Learn more about the weeknotes concept here.

    One Thought: What happens when an AI launch is half-baked

    (Reposted from LinkedIn) If you launch #AI into your government operation without thinking it through, this is the kind of story you can expect:

    ‘AI-Mazing Tech-Venture’: National Archives Pushes Google Gemini AI on Employees

    There is no need to deploy #genAI as an early adopter unless you have a large ML-friendly data set that can be mined for BI-style insights that you couldn’t get with traditional methods.

    Should you experiment at small scale in lab conditions? Sure, if you have discretionary budget. But launching an LLM at employees (maybe to declare some kind of “leadership” victory?) will look desperate and may cause a lot of trouble.

    This is doubly true if your agency doles out facts to the public and adds AI into the official service chain.

    And just politically speaking, the U.S. National Archives and Records Administration (NARA) should probably be one of the last to deploy AI, not the first.

    Do we really want AI bots addressing public questions of who shot Kennedy, or did Trump illegally withhold documents from NARA itself?

    Five Notes

    1. It’s already been widely publicized and praised, but in case anyone missed it, the Digital Service Network at the Beeck Center released their Government Digital Service Team Tracker this week and it’s a fantastic resource for finding other digital service teams across the country and explore their sizes, structures, and roles. And yes, our very own GX Foundry is in there.
      • BONUS: That team tracker is just one resource sitting inside the Digital Government Hub, a new centralized database of resources to help out government digital teams as they work toward their myriad goals. It’s a stunning achievement and deep resource.
    2. Stay tuned to our GX Foundry website for 2 new positions we will be posting next week. We will be adding a Digital Product Owner and a Digital Communications Strategist to the team that is leading the One Franklin County project. Both will report to hiring manager Sarah Gray.
    3. HBR re-released a podcast episode titled How to Make Better Hiring and Firing Decisions that’s really solid. Recommended advice.
    4. I can’t remember how this piece got to me, but it was new this week and it’s (sadly) very, very applicable to my broader organization. Nice, nice, very nice. The piece is all about how an excessively “nice” organization culture can—surprisingly—destroy psychological safety and dissolve trust within teams.
    5. The Rework podcast this week tackled a favorite theme of theirs: saying “no” to most requests and options. As the 37signals guys note in Say No by Default from experience, saying “yes” is deceptive. Because when you say yes, you’re actually saying no to hundreds of other options and possibilities. Saying yes is a cop-out, to get out of tough conversations about what really matters and making tough choices.

    One Video: Beedeebeedeebeedee… What’s up, Buck?

    I’m old enough to have seen Buck Rogers in the 25th Century (1979) on broadcast TV. No cable. No DVR. No VHS. Just broadcast television, with the antenna and everything. I even had a model of the Buck Rogers fighter ship. This was just one show that came out of an explosion of sci-fi after Star Wars (1977) revived the genre. Battlestar Galactica (1978) was another popular example, and I saw that show’s feature-length pilot in a movie theater before it made it to the small screen. It was a wild time.

    This show did not end well. It also didn’t start well. But as a little kid, I didn’t know that, and with only 4 broadcast stations to choose from, I was starved for media (compared to kids today). I was also a little transfixed as a kid because Mel Blanc—the voice of Bugs Bunny and more—was obviously voicing the robot Twiki. That blew my little kid mind.

    I had fun taking this tour through a ridiculous bit of TV history. For the kids out there, this stuff is both laughable and cringe, and it shows just how far we’ve come. Compare this cash-grab to the likes of The Expanse (2015). That’s a huge leap.

    https://www.youtube.com/watch?v=3jGD6r0z_O0

    Five Laughs

    One Photo

    I didn’t realize you could have something like a “Saguaro Forest” in the desert southwest, but indeed you can. This was on the road to Mount Lemmon, northeast of Tucson earlier this month. The heat wave hitting the area was a little unwelcome during our visit, but as the days have turned a bit cold and grey this week… hmm… that heat wave doesn’t sound so bad.

    https://digitalpolity.com/2024/10/19/2024-weeknote-42-2025-hiring-is-around-the-corner/

    #AI #Beeck #BeeckCenter #BuckRogers #DigitalGovernmentHub #DigitalServiceNetwork #DSN #firing #funnies #genAI #Gx #GXFoundry #HBR #Hiring #NARA #nice #no #Saguaro

  2. September 2 – 8

    My “weeknotes” capture events, thoughts, and other items from the past week, mostly focused on work.

    [1] Your ever-so-slightly-different writing style probably looks like you faked it with an AI tool (when scanned with AI tools)

    We’re clearly on the back side of the AI hype cycle, rapidly sliding down that heady peak in the chart into a trough of despair (for a bit, anyway). That’s good. Because I just learned about another problem: AI tools cannot accurately detect AI-generated writing. (Which is funny, because I can.)

    I was tipped off by this Bluesky exchange:

    Thank God I’m not in college anymore. I’d get flagged as AI all day long. Meanwhile, Dr. Williams also posted this exchange, explaining the problem a bit more:

    Bottom Line: AI tools have zero intelligence. They just have math. The LLM-based tools are good at statistically estimating which words likely come next in a sentence from macro- and meso-language indicators based on an ingestion of a huge corpus of text. But the source text matters tremendously, both for producing new statistically-likely passages and for detecting passages that may be statistically-generated by other LLMs.

    And may the AI gods help you if you’re creating content for U.S. and English-speaking European consumers but you aren’t originally from this part of the world, as noted in this WIRED piece: “AI models were mostly trained on data from and for Western markets, and therefore can’t really recognize anything that falls outside of those parameters.”

    I remain hopeful that there will be tightly-targeted uses for generative AI in a few areas. But I’m glad to see NVIDIA’s stock start to fall back to Earth and articles like these start to get recognition. The breathless “age of AI” will really just be some nice innovations around the edges, not a total re-think of technology.

    NVIDIA’s stock has been up over 2,200% in the last 5 years, with most of that coming in the last year due to expectations around their hardware-level participation in the generative AI market.

    [2] Miscellanea

    • I was surprised to see the likes of Bloomberg take on the biggest player in local government software this week, in a piece titled: How Local Governments Got Hooked on One Company’s Janky Software. Bloomberg released this blockbuster report on September 5, pointing to lots of Tyler Technologies problems, like a checkered history of failed deployments, technical screw-ups (some of which put thousands of people in jail without justification), and lots of related lawsuits. I had no idea local government software woes could get national attention.
      • I have a lot of thoughts about this Bloomberg piece because Franklin County just killed a multi-year, multi-million-dollar contract with Tyler this summer.
      • I’m debating whether to share those thoughts publicly. But don’t get too excited—my comments are about software strategy, not the specifics of our experience with Tyler.
    • As noted previously, we are hiring an Application Developer for our GX Development team. And in less than one week we completed 17 screening calls—a record. I was in some of the calls, but it was Eric Nutt that hit this high-water mark. Really remarkable achievement. We’re moving fast due to the weird process we use for hiring. We need a finalist by 9/27 or we can’t hire until December.
    • Listened to a great podcast episode this week—Coaching for Leaders: The Habits That Hold Leaders Back, with Marshall Goldsmith. Goldsmith is the guy that wrote the bestseller What Got You Here Won’t Get You There: How Successful People Become Even More Successful and he has plenty of good advice.
    • Due to an unattended package found outside the courts building on Thursday morning (which was perceived as a potential threat) most employees were remote on Thursday, so we held our 5th Hack Your Bureaucracy book club via Teams and it was okay. Got to use the Breakout Rooms feature for the first time. Still prefer in-person, though.
    • Joined the Digital Service Network‘s Chief Digital Service Officer call on Thursday afternoon and learned the State of Maryland is running into the same problems we are with website consolidation, rationalization, and long-term ownership. It’s good to know we’re not alone, and neither are they! Government across the country are dealing with this past proliferation and trying to bring it back together to build better services for residents.
    • And Thursday night I made a final trip to see the Columbus Clippers with some colleagues from the office. Summer is truly at an end.
    • Quickbase posted videos from their national Empower24 conference to YouTube this week, including the presentation that included yours truly, Eric Nutt, and Luke McCormac talking about our Unclaimed Funds software project. You can see it below:

    https://youtu.be/oSae3eGtKBY

    [3] Watch This

    I hope Paramount keeps this on YouTube! Starring Archer’s H. Jon Benjamin it’s an exploration of what happens when Starfleet doesn’t do a great job hiring scientists. Anyone who’s been in a staff management role will cringe, and laugh, at this scenario. Stay tuned to the end for the hilarious Tribbles cereal commercial.

    https://www.youtube.com/watch?v=kkDLgf2em68

    [4] Internet Funnies

    https://digitalpolity.com/2024/09/09/2024-weeknote-36-here-at-the-end-of-summer-ai-cant-see-ai/

    #AI #baseball #Beeck #Bloomberg #bookClub #ColumbusClippers #court #courts #DigitalServiceNetwork #DSN #HackYourBureaucracy #Hiring #InternetFunnies #Leadership #LLM #Maryland #NVIDIA #podcast #quickbase #software #StarTrek #strategy #tribbles #Tyler #TylerTech #TylerTechnologies