s09e13: Mist, everywhere
0.0 Context Setting
It’s Friday April 2, 2021. I’m back in Portland after taking our kids to the coast for a few days and it’s warm and sunny.
We’re going to go out and probably play tennis this afternoon.
It’s now been something like two weeks since I’ve tried to delete this publication on Substack and it still just errors out.
1.0 Some Things That Caught My Attention
A SAFE COMBINATION: SECURITY AND YOU
hubot-security-posters is a node.js module that returns a random NSA security poster from the 1950s-60s. If you just want to see the posters, they’re stored as 138 PNG images in this github repository.
Caught my attention because: FOIA requests, internal communications, national security, poster design.
Dates and Old News
A couple days ago, Jacobian asked for ways to deal with old, outdated blog posts, Simon Willison answered that he’d make sure the date is visible on them, I remembered that The Guardian has a good model for this - they include a highlighted yellow banner at the top of an article that says something like “This article is more than 3 years old”, which is all linked in my head to this post from ProPublica last week about ProPublica’s new article page design… which I was pleasantly surprised to see mention the involvement of agency Happy Cog and Jesse von Doom.
Caught my attention because: design and delivery of information online, combating misinformation, providing context to information, thoughts about contextual cues as to information metadata (e.g. what “old” newspaper paper looks like, and the loss/change of that contextual information in the shift to digital), everyone loves a good design system, Jesse von Doom is working on substationHQ, a super interesting platform for paid newsletters, and I have good memories of hiring and working with Happy Cog on the web design (how quaint) for that long-ago ARG, Perplex City.
ARM servers
Honeycomb is a company that produces observability tools, that helps people understand how their systems are working. They recently posted about using Graviton2, Amazon’s custom ARM-architecture processor designed by wholly-owned subsidiary Annapurna Labs. Jon Masters at RedHat replied, asking about the remaining 8% of compute that Honeycomb are running on x86 still (too many distinct/heterogenous workloads to migrate, a bunch of random stuff), with the view that in a few years, people would keep x86 “systems round just to run the few things they don’t bother to migrate”. Which on the one hand, I see in some infrastructure, but perhaps not in the vast majority of non-SaaS infrastructure. I mean, have you seen how old some systems are out there?
Caught my attention because: triggered my Sun dtrace pattern recognizer (cue instant Joyent flashback), earlier thoughts about being interested in methods of figuring out/understanding the stack/infrastructure of large systems, processor architecture, underlying infrastructural change and how quickly/slowly people think it might happen.
Why does it always have to be planes?
In a thread about game development (“Man, 90% of the terror of game development is just stuff you have not done before one (1) time, huh”), Tom Forsyth added this analogy about what it’s like to make games:
Jumping off the cliff, building a plane on the way down. But not a normal plane. We want a new and different plane Tweet
Caught my attention because: the plane analogy is one that keeps coming up with software (completely replace a plane while it’s in flight without crashing it, etc), and thinking about why it’s always planes? Is it because they’re relatable large, complex systems? Because they crash, and there’s an easy cognitive link in not wanting this to crash? Is it the complex procedure of flying them? But I do love the conflation of jumping off a cliff, which feels so wonderfully agile-ish (as a user, I want to get to the bottom, so that…) and having just enough planning to know you’re going to jump, and then adding on new and exciting requirements (that can be self-inflicted or externally imposed), which results in things like “we want a new and different plane”.
Fanfic set in the Iain M. Banks Culture universe
I vaguely remembered reading fan fiction about what might happen if Iain M. Banks’ Culture encountered… something from another fictional universe, and managed to find it! Here’s the favorite Culture fanfic I found: (on archiveofourown.org, of course)
- That time Cheradenine Zakalwe recruited Q. (No, not the Star Trek Q. The one with the Aston Martins).
- That time someone crossed the Culture with Terry Pratchett, introducing the GNU Man In The Funny Black Hat.
- That time the GCU Grey Area had its first Special Circumstances mission.
- The Culture vs The Empire, in Listen, Strange Space Wizards Sitting In Temples Distributing Laser Swords Is No Basis For A System Of Government
- And the one that I remembered that started it all, The One Where The Culture Explores The Universe Of Warhammer 40,000, which, upon re-reading again, helped me realize I really wasn’t that into Warhammer 40,000.
Caught my attention because: I mean, it’s The Culture. And also, mashing up fictional universes together is pleasing to my brain in very distinct what if ways. Anyway, here’s an… essay… about the Culture from The New Atlantis.
An affordable solution that offers a widespread, revolutionary solution
cw: guns, shooting, killing.
Via The Drive, the Pentagon has hired Smart Shooter to develop a smart sight for U.S. special operations forces American Special Operators Set To Get New Advanced Computerized Sights For Their Rifles.
It’s a live computer vision system that puts a reticle on a target, helpfully selects (human?) targets, and “quickly calculates the optimal point of aim, even if the target, such as a small drone [hmmm], is moving.”.
There is, of course, a product video on YouTube.
Caught my attention because: oh, wow, yeah, computers and software and killing. Exasperated Sarah Connor sound.
Oh wait, this also caught my attention because the U.S. Army and Microsoft recently announced that they’re going to move to the _production phase of the Integrated Visual Augmentation System (IVAS) [Microsoft]. Here’s a November 2020 piece on IVAS, from the U.S. Army’s blog.
I imagine that the system might one day be able to identify the size and match of a biker’s clothes, boots and motorcycle.
Good Dog
Oh right, here’s something nice: someone used the YOLOv3 object detection model trained on the COCO dataset to recognize dogs when they go past [Github code for Dog Detector] which is all well and good, but then much better if that same person goes the extra step to hook it up to automatically shout compliments [Youtube video].
Caught my attention because: it’s fun, whimsical, uses off-the-shelf hardware (a Raspberry Pi) and off-the-shelf ML (YOLOv3, which also has a silly acronym, named for You Only Look once), it feels like the kind of thing Google wanted to see people do with its AIY kits (Do-it-yourself artificial intelligence), in particular its $99.95 AIY Vision kit (which itself is powered by… a Raspberry Pi).
Actually, what also caught my attention about this is the further reminder of the ubiquity of the Raspberry Pi as a sort of proto-computational dust and its history. You know what, let’s do this…
Mist
Another example of how my brain works: I saw some Intel ad about how it was going to bring its new/cloud chip technology to the Edge, which is another way of saying “we need to grow our business because maybe the datacenter market will only get so big, so let’s add more compute closer to the outside bits of networks”, and honestly my first thought was huh, when the Cloud reaches the Edge and compute becomes pervasive, we should call it Mist.
Anyway, look. (Disclaimer: armchair uninformed theorizing) There’s one view of how compute is everywhere, which is: hey, there’s lots of microcontrollers all over the place, from your dishwasher to, I don’t know, toasters or whatever. Or your car.
The ubiquity of microcontrollers (you could buy an 8-bit one for about 32 cents in lots of 1,000 in 2018 [Wikipedia] is partly down to stuff like Moore’s Law and wanting to do stuff with all those fabs lying around after the big expensive processors in computers move on to the next manufacturing process.
So, you know, there’s already pervasive computing, but the more recent (ha, more like at least 20 years) view is that the new kind of more interesting pervasive, or ubiquitous computing would be the kind that can sense its environment and perform some sort of output and is also networked.
Raspberry Pis are super interesting in this respect! Just look at this list of Reasons Why They’d Catch My Attention:
- They’re all over the place and people keep buying them for things like running Pi-hole, a cheap quick way of blocking ads using a little computer
- They’re British, which triggers a whole bunch of BBC Micro memories and associations: I mean, could you imagine, a government investing in a national program to kickstart use and understanding of computers through a state-owned broadcasting corporation?! What sort of insane public service would this be?! What could such a broadcasting corporation turn into, with history like that, with the advent of the internet?! (wait, don’t answer that)
- They’re a registered charity [Raspberry Pi - About], which my gut tells me has a lot to do with their ubiquity.
- It’s super interesting to think about the Pi then in comparison to something like the OLPC project.
I mean, there’s still a bunch of difficult stuff to do with using them, but they’re cheap they range from $5 to $75 for just the boards, and you can get the Pi 400, which is the one with a keyboard, for only $70.
I just love the experimentation. There’s the ubicomp in terms of all the corporate stuff, all your regular IoT. And then there’s “hey, I built this thing that recognizes dogs and then says compliments to them”, that maybe cost $100 or less. And, you know, it comes out of Cambridge, its original purpose was to further the study of computer science at the school level, a pretty much literal replay of the original BBC Micro strategy.
So. Mist. There’s corporate mist. And then there’ll be Mist From The People, Doing Awesome, Fun, Useful People Stuff That Apparently Isn’t Worth Doing Because It Isn’t Profitable.
Oh, and it’s been about 10 years since the foundation started, which totally fits my confirmation bias of most big things take about 10 years.
From the archive
I just remembered the time I wrote about Stages of [digital] Transformation from, huh, just about 4 years ago, it looks like.
Caught my attention because: the plane metaphor, probably, and how, well, digital transformation.
OK, that’s it for today: about 1,800 words in about an hour.
How are you doing?
Best,
Dan