February 26, 2014

Episode Twenty Five: Structure; Humble, Direct and Curiously Cautious; The Scale Of Things, Risks

Rambling again today, I'm afraid. Do let me know what you think of this one, because it feels somewhat unclearer to me than some of my recent episodes.

1.0 Organising Structure

I'm still preoccupied with the idea of there being nascent large-scale functional structures on the internet in the same way that neuroscientists study the structure of the human brain and discern individual units (such as they are) of functionality. There's so much to unpack here: what are the structures that are being built, for starters, never mind for what reason they're being built by us worker ants. 

What other sort of parallels with specieisation, specialisation and evolution are possible in looking at the growth of the network? As we move to distinct services being made available as cloneable code and easy to spin up cloud instances, and defining interfaces for those services (e.g. the trend for *-as-a-service), it's easy for us to see the "computer vision bit" that gets shunted some data and the "hearing human speech bit" that gets shunted other data. There's no real co-ordination, no one entity saying "let's make an internet that understands the world" but instead lots of little sub-problems that are busy being solved for whatever reason (praise the market). Instead of vision systems evolved to detect potential predator threats, like peripheral vision motion detection, we have systems designed to detect trending topics. 

We know what kind of structures get produced when biological evolution and replicators are present in a physical environment. Actually, that seems a bit like a silly statement: we don't know at all. We happen to know a lot of implicit things because we're the product of that environment. But the type of structures that we're erecting as infrastructure across the web that bear some sort of macro resemblance to what has been left in our brains are just *weird*. We have bits of brain that are good at detecting edges or faces or decoding the mush of sound that enters ears shaped *just so*, but what are we making computers understand? Bits that recognise "interestingness" in images based on the droppings that we hairless apes leave around them? Bits that that clickstreams and try to work out relationships between them? If we're the product of a physical embodied intelligence, then if you tell me that you know how to relate to or understand and intelligence that is embodied in the warp and weft of data, I have no idea what you're on about. I did have a thought about these functional structures not quite yet having reach out into the physical world and realtime feedback as to those actions, but I suppose autonomous drones are bringing that future closer. And also, there's my physical-instantiation-privilege leaking through, prizing the material world over the one made of data. 

So, in that tradition of the lazyweb, I would like this: what are the nascent neurological-analog structures of the internet? Where's the equivalent of Broca's region or an off-the-shelf visual cortex? Is the profusion of key-value stores the equivalent of laying down long-term memory? Put it this way: the internet has gone from not being able to recognise faces to being able to do within thirty years. I'm not necessarily a Kurzweil adherent, but you've got to admit, that's not bad going. 

2.0 Humble, Direct and Curiously Cautious

That's how Robin Sloan described[1] the team at BERG and the way they announced their latest project, Cloudwash, on how their product, BERG Cloud, could integrate with an off-the-shelf washing machine.

It's worth watching the Cloudwash video[2]; it's a particularly British and understated film explaining what might be so compelling about an internet-connected washing machine and the process that goes into designing such an object and one that you could see yourself using in your home. 

BERG's video doesn't fall into the trope of Jony Ive Explains Things Eagerly In A British Manner, but Sloan has identified something uniquely British and lacking in Valley boosterism in the way that the team at BERG patiently explain what it is they're trying to achieve, why and how. 

In the way that we have a fairly identifiable consensus of what makes a West Coast startup (and I would argue that, in the grand scheme of things, an East Coast startup is practically indistinguishable) I'd love to see some sort of Dogme 95[3]-esque backlash or reaction to the Established Way Of Doing Things that takes it cues from a more British sensibility.

It helps, of course, that the team behind BERG embrace their Britishness, their shop-coats (not lab coats, mind), and their emphasis on Getting Excited and Making Things. Where the Valley stands for hype and overstated rhetoric, it's instead the British and BERGian way to be self-effacing and humble. Where America built itself the Lean Startup manifesto[4], what sort of manifesto might emerge out of a particular British sensibility of being humble, direct and curiously cautious? 

This is not the Britain of empire and colonialism, more a Britain of shop floor tinkerers and engineers who were doing things as hobbyists and accidentally invented entire industries, not out of some sort of copy-and-pasted love of money, but of material and technique and craftsmanship. This isn't that particular American sensibility of "artisanal" that's been sprouting lately, but a different kind of retro. It's an open secret[5] that the name of BERG is a direct reference to the fictional British scientist Quatermass[6] and his British Experimental Rocket Group, and while there are some things to say about the culture at BERG (one could say that it could be a little testosteroney at times), there's no denying that Messrs Schulze, Webb and Jones forged a distinct atmosphere and culture there, one refreshingly different from the dominant West Coast narrative.

I said on Twitter that one of the irritating things about BERG is that when you see their documentation, they make what they do seem easy, and in implication make everyone else appear to be quite dumb in comparison. One look at the Cloudwash demo product is enough to convince you that yes, there's actually value in connecting a washing machine to the internet once you sit down and think about it properly. Gruber talked about this in his post about working backwards to the technology[7], and it's perhaps this aspect that BERG shows: their focus on design and technology in the service of design. On the one hand you could look at a company like BERG and its cloud product positioned as the glue for hardware manufacturers looking to make internet connected products as a solution looking for problems. What BERG have figured out, though, is that no one's really in the market for a solution that is also looking for problems: and that they have to lead the market to the solutions that BERG Cloud enables. It's a case of timing (whether good or bad) that the position they're in is that the best way of doing that is by inventing the products themselves. It used to be the case that BERG would simply do video demos - one criticism of them was that while they were great at design fiction, they were even better at creating shiny videos. Cloudwash is anything but that: a real, physical prototype that works because how else to sell the utility of BERG Cloud the product? It's this focus on the job-to-be-done, the working back from the problem and illustrating *how* BERG Cloud fits into the solution that's interesting and in a deep way, at odds with a stereotypical Silicon Valley technology-focussed startup culture. 

But: and with a heavy heart, I have to admit, it feels like Britain is the outlier with companies like BERG. The dominant narrative is not one of humility, directness and curious caution. It is instead one of brash optimism and forging ahead - the inventing of technology for technology's sake. I wish BERG all the best and hope that they aren't relegated to the role of an externalised, barely paid-for research and development outfit whose ideas can be copied without credit or remuneration by the industries they're hoping to improve. They deserve better than that.  

[1] https://twitter.com/robinsloan/status/438474116551548928
[2] http://blog.bergcloud.com/2014/02/25/cloudwash/ and https://vimeo.com/87522764
[3] http://en.wikipedia.org/wiki/Dogme_95
[4] http://theleanstartup.com
[5] http://berglondon.com/blog/2011/02/21/our-experimental-rockets-are-our-people/
[6] http://en.wikipedia.org/wiki/Bernard_Quatermass
[7] http://daringfireball.net/2014/02/working_backwards

3.0 The Scale Of Things

The sale of WhatsApp to Facebook brought into public eye the scale of internet companies again. The grammar when looking at consumer internet companies is to speak in DAUs and MAUs - daily and monthly active users - as a proxy for "engagement" and the theoretical maximum ability of a company to monetise that particular audience. It's interesting to think about what these large numbers mean and imply - not the purchase price - but the user base. So, here is a list of numbers.


4,700 - Viewers of the 2008 Suer Olympics (watched part of coverage)
1,700 - Servings of Coca-Cola per day
1,460 - Rail network passengers, United Kingdom, yearly
1,358 - Movie theater attendance, United States, yearly
1,351 - Population of China
1,237 - Population of India
1,230 - Facebook monthly active users
815 - Airline passengers, United States
757 - Facebook daily active users
556 - Facebook daily active users, Mobile
530 - Watched the first man on the moon
450 - WhatsApp monthly active users
359 - Theme Park attendance, United States, yearly
313 - Population of the United States
260 - Viewers of the final game of the 2006 Fifa World Cup
254 - Total number of cars in the United States
241 - Twitter monthly active users
198 - Population of Brasil
184 - Twitter monthly active users, Mobile
131 - Sports event attendance, United States, yearly
127 - Population of Japan
116 - Delta Airline passengers, combined system
110 - PlayStation Network users
100 - Minecraft registered accounts
81 - Population of Germany
75 - Instagram daily active users
65 - Steam monthly active users
63 - Population of the United Kingdom
61 - Sina Weibo daily active users
48 - Xbox Live users
48 - Cable subscribers, United States
40 - Netflix global streaming subscribers
38 - Population of the State of California
23 - Population of Australia
21 - Washing machines sales, European Union, yearly
21 - Comcast (pre-Time Warner merger) total subscribers
14 - Population of Shanghai
8 - Washing machines sales, United States, yearly
0.7 - Population of the City of Detroit

(There was a book I had growing up, The Guinness Book of Answers, a sort of printed-out one-volume Wikipedia that had everything from lists of Gods through to scientific units through to geographical information, a sort of CIA factbook on steroids. This list reminds me of that book).

4.0 Risks

A short last one, this. Since my days at university before the 21st century, I've been on-and-off reading the RISKS digest[1], a moderated mailing list highlighting risks to the public in computers and related systems. Between it and Bruce Schneier's Cryptogram[2], the two mailing lists deftly set out exactly how alarming it is that so much of our lives rely on software that can barely lay claim to the term 'engineering'. Apple's iOS and OS X SSL bug brought these two lists to the fore of my attention again, and a discussion with a friend as to the state of Federal computer system procurement (a particular conversation about the particular extent to which healthcare.gov was a cockup - and it's worse than you can imagine) reminded me that it almost seems like we have a clinical inability to see the complexity in systems. I don't see it as merely something that's lacking political will, more something we explicitly decide to pay no attention to. It is not outside the realm of possibility to engineer reliable software: an oft-cited article is They Right The Write Stuff, a Fast Company piece from 1996. In the case of the US government, the interstate highway system instituted in the 1950s established both a standard for the building of national network infrastructure as well as funding. The roads did not break. They carried the traffic. It is strange to me that we do not hold software to the same standard, at least for one reason being that the complexity is hidden behind the veneer of interface. I do not know what it would take for software to be taken seriously.

[1] http://catless.ncl.ac.uk/Risks/
[2] https://www.schneier.com/crypto-gram.html
[3] http://www.fastcompany.com/28121/they-write-right-stuff


That's all for today. Happy Wednesday, and see you tomorrow. As ever, I appreciate notes and replies from you, even if they're short ones.

Best regards,