November 01, 2016

s3e33: DNS Fan Fiction; 100-watt-hour-Trucks

0.0 Station Ident

2:44pm on Tuesday, 1 November 2016 in Oakland, California at the Code for America Summit listening to the dulcet tones of Tom Loosemore talk candidly about what worked and what didn't work at the UK's Government Digital Service. 

If you're at the Code for America Summit in Oakland Wednesday 2 and Thursday 3 November and you'd like to say hi and talk about, well, anything, but probably more likely to do with digital transformation in government, then drop me a line. I always like meeting interesting people and chatting. 

1.0 DNS Fan Fiction

There was a ridiculous story[0] in Slate the other day about a mail server talking to another mail server which normally wouldn't be news, but is news (in America, at least) because it has something to do with a) Russia, and b) Donald Trump and c) oh yes, that election. The story is worth reading for entertainment value only, I'd say. 

In the spirit of that Slate piece, here's some DNS fan fiction, a version of which I idly wrote on Twitter this morning[1]:
Strands of binary ones and zeroes course through the digital nervous system, also known as the D.N.S., subverted by criminal syndicates. This story starts with Akira - one of the most notorious cyber scientists, known on the darknet "Reddit" as a white-hat hacker, thanks to his digital avatar that sported a white hat.

When he reached out to me, what Akira told me was simply too stunning to be true. He and his cyber colleagues, he said, had uncovered evidence that high-powered computers in server farms belonging to and associated with Donald Trump's business interests, were using military grade encryption to communicate with each other, and also unknown third parties - potentially in Russia. 

How did Akira and his collaborators know? The answer lies in an esoteric branch of mathematics called Prime Number Theory. Prime number theory isn't even taught in schools anymore: in off-the-record conversations for fear of criminal investigation, computer scientists have told me that some prime numbers are so dangerous that they are illegal to even write down or say out loud. Nowadays, prime numbers are used by organizations like the National Security Agency to send unbreakable messages that help protect our nation's security. 

Could it be possible, I wondered, that someone at the National Security Agency was leaking secret prime numbers to the Trump campaign to rig the election? To answer this, and to learn the scary truth about the dark forces hijacking the arcane digital nervous D.N.S. system, I had to delve deeper.

 By now, the hacker darknet site "Reddit" has penetrated the public consciousness. Even amongst Reddit users, though, hardly any know of a more secret network. Originally founded as the Universal Security Enforcement NETwork in the 1980s, USENET, to use its sinister name, is a dangerous, illicit cyber location. A haven for cyber vigilantes, USENET is a place where outcast computer scientists are free from rules and regulations to experiment.

How dangerous is USENET? The recent outage that affected the Internet in fact came from an attempt by individuals linked to USENET to create something never before seen or attempted: a robotic network intelligence, also known as BOTnet. The BOTnet was a combination of cutting edge, ethically dubious "deep learning" techniques as well as millions of sophisticated net cameras. And now, I'd just learned from Akira and his contacts that the Trump Campaign - by using the same secret prime numbers that power the internet's digital nervous system - was linked to this same attack on the Internet.

The revelation that the Trump Campaign is using prime numbers to attack the Internet and to rig the election is not only terrifying but also raises questions. Where are these prime numbers coming from? Are they really coming from the N.S.A. and if not, how is the Trump Campaign getting them? 

This whole investigation was becoming fractal, like a jigsaw puzzle with a million pieces. I needed a guide, a native to the cyber underworld, someone who would be willing to teach me the secrets like the "routing protocol" I'd need to understand to navigate the dark nets. Fortunately, my earlier contact Akira was able to recommend someone who I would only ever know by the pseudonym of "Hiro", a reference they assured me to an underground cyber text from the 90s. 

With a chuckle, Hiro sends me on a 'quest', their way of playing with me like a cat does with a mouse, to make me jump through hoops to show that I'm serious about following them on this journey. 

It's late. I've been chasing this rabbit of how deep the Trump Campaign is subverting the Internet for days now, with little sleep. I've learned more than I want to, but it's not enough. What I have learned though, is this: our world isn't ruled by power or information anymore. It's ruled by incomprehensible ones and zeroes. Could it be, I worried and wondered, that the Trump Campaign - through actors and assistance unknown - had effectively stolen and subverted the meaning of the very digits that control our world? And what then of the rumors I had started to hear - the most sinister ones, advanced by computer scientists strictly off-the-record - that the DNC and GOP had in fact struck an underhand deal: half the Internet - the ones - to the DNC and Hillary, and the other half - the zeroes - to the GOP and Trump? Was this what Trump was trying to say when he was saying the system was rigged? 

I had to follow up this last rumor, at the very least. Sources in the Trump Campaign would later deny any knowledge at all of such a deal to carve up the Internet. It would be inconceivable, staffers said, that Trump would get the zeros and Hillary would get the ones. "Trump," they would say, "is not a zero."
[0] Was a server registered to the Trump Organization communicating with Russia’s Alfa Bank?
[1] Dan Hon on Twitter: ""Strands of binary ones and zeroes course down the digital nervous system, also known as the D.N.S., subverted by criminal syndicates""

2.0 Inside Baseball

I'm sorry, but I've been thinking a bunch about Apple's recent product announcement of the new MacBook Pro. For what it's worth, the things-that-have-caught-my-attention about this are less about the specifics of what Apple did or did not put in their new pro laptops, but more about what "computing" looks like these days. So, some thoughts that are coalescing:

* I think we can agree that desktop computers are a) not a growing market, b) a shrinking market as user needs are better met by mobile computing (whether phones, tablets or laptops), c) probably one that won't disappear to zero, but one that will consolidate and re-baseline.

* Why? I think ultimately what we've seen Apple do is try to optimize against a) battery life (which impacts mobility), b) thinness and lightness (which impacts mobility) and c) performance. As they say, if you have three, you can only pick two. 

* Just like the way that mobile phones have displaced single-purpose cameras as the devices that people use to take photographs (ie: it doesn't matter what *kind* of camera you have, the best camera is the one you have with you), the best computing device *most of the time* is the one that you have with you. Computing, it turns out, is something that's useful wherever and whenever you are, and turns out to not just be useful when you're tethered to a mains outlet. 

* Thanks to Moore's law *and* the network, *most of the time* we have roughly enough computing power to do roughly most of the computing tasks that people need to do in an energy envelope that also sits inside current battery life capability (10 hours per full charge).

* There's a real constraint - one that's not physical - which is that in the U.S., the FAA says that the biggest lithium-ion battery you can take on to a plane and still use is one with 100 watt-hours. So! If you're talking about a "mobile" device, you're kind of limited to that energy envelope. Everything you do - if you agree that the most useful computing device is the one you have with you - needs to fit inside that energy envelope. 

* So! The big question is - what computing tasks *cannot be done* reasonably (in terms of performance requirements, etc) inside 100 watt-hours worth of energy envelope? Your 100 watt-hours (Apple's newest MacBook Pro has a 76 watt-hour battery that reportedly hits 10 hours of usage). Again, for *most* computing tasks (e.g. web browsing, web-based transactions and "light" content creation), we're more preoccupied now with *efficiency* - ie performance-per-watt, the reason why Apple switched from PowerPC to Intel in the first place! - rather than sheer performance. 

* This comes back to the crux of it: performance is an issue of comfort in terms of getting the job done. I *can* edit 4k video on a brand new MacBook Pro, but the fundamental point is that my capability in terms of editing video (or any other kind of task that benefits directly from high-performance, high energy-requirement compute, whether CPU or GPU) *must* fit inside 100 watt-hours. My CPU is limited and my GPU is limited. There are, of course, CPUs and GPUs that fit outside that envelope and are way more performant. How many people *need*, and are willing to pay for that performance? 

* I think the point is this: there are *some* people who need that performance, but in the context of the entire market the relative percentage of users who *need* 100 watt-hour+ performance is tiny compared to *everyone else* who is satisfied by sub 100 watt-hour performance. 

* This, of course, goes back to Steve Jobs' remark about computers and trucks. I'd say now that "trucks" are 100 watt-hour+ perfomance envelope computers. *Everything else* is your regular MacBook or iPad Pro or iPad or iPhone or whatever. To go Apple kremlinology right now, the issue is that Apple's 100 watt-hour+ performance envelope computers are *no longer competitive* and don't make the most of, nor are true to having that energy envelope. 

* Actually, I kind of like that way of thinking about things. Apple, *even* in the 100-watt-hour+ performance envelope *still* prioritize thin-ness and form in terms of industrial design. In this case, I'd say that Jony Ive's design direction is crippling Apple's ability to lead in that area. The iMac series - which is the only reasonable option for most people in the 100-watt-hour+ performance envelope *still* use mobile-style components due to the industrial design requirement that the computers *look* good. But if Apple is to say that design is *how it works* as well as *how it looks*, then how it works *also*. From this point of view, the Mac Pro's industrial design was a major mis-step in terms of its lack of expandability that leans in to the 100-watt-hour+ envelope (ie: you've got a desktop computer, it's perfectly fine for you to want to put an energy hungry newer, better, faster GPU in it... but you're *not allowed* because it's built in. 

* I think the point here is that in the way that the computing truck (100-watt-hour+) is a) a real market, b) not a growing market, c) but a market opportunity nonetheless, Apple has shifted from having made viable trucks (the Cheesegrater Mac Pro) to being (understandably) excited both from a financial and change-the-world-perspective of solving "transport" for everyone. I think part of the issue and blowback from the peanut gallery is more around the fact that macOS is tied to the hardware. There are people who *want* to use macOS but don't have a choice about a viable, realistic truck to use anymore. At least, they don't because they have no idea what Apple's plans are in the truck space. 

* Long story short: market segmentation and the size of the market for truck-type computing devices. I'm less interested in litigating the point as to whether it's *worth* Apple to make a truck (they should), but more that as the digital landscape matures (forgive me for that expression), we're able to identify new specific market segments that are viable. 

3.0 Your Partner's An A.I.

Robin Sloan wrote me a note in response to my thought the other day about deep-learning systems that work with us instead of replacing us and the whole centaur model; I'd written about how excited I was about AI as prompts, using his auto-complete science fiction text editor. Robin's point was that he thinks the opportunity is less AI collaborating with us on *existing* formats (ie: an AI assist for novel writing), but that AI can create *new* formats and fill them in itself. The way Robin put it was "new forms for new channels", which I'm super interested in but I'm not entirely sure about. I think the collision here was the announced shutdown of Vine. It was a human decision that created the Vine 6-second video constraint and then once the format had been chosen - or curated - once that design space was dictated, *then* filling in that 6-second format is something that, I think, an AI could be super interesting at. 

So my point is that, *yes* I think AIs can be super interesting in terms of coming up with potential formats but perhaps the issue is that AIs do possibility generation *better* than we do. Something still has to pick and choose! Which we would do, because hey, we still have to do something. But then! We can hand it over to the AI and say, hey, we chose 3-second looping gifs as a format from the selection you presented, now make shit up. 


OK, all done. 4:34pm and I'm being super rude by not paying attention to Mr. Loosemore.

Remember, if by any chance you're at the Code for America summit and would like to chat, drop me a line. And always, I love replies and always try to reply to them.