How do we build a self-conscious machine?

I write a weekly newsletter Exponential View, exploring the intersections of technology and society.

This week’s issue looks at trust in the age of platforms and the future of mobility. Subscribe to receive Exponential View every Sunday.

DEPT OF THE NEAR FUTURE

🤔 Building a self-conscious machine. Fascinating long-read by Hugh Howey on how we might build self-conscious machine. It may need a theory of mind, internal storytelling, and messy modules rather than engineered precision.

🚀 Blockchain is a new digital substrate that makes old models of governance, development, and business obsolete, argues Angus Hervey. This good survey on the promise of distributed ledgers argues that:

the combination of a shared, open database with permissionless innovation and an incentive system that prevents winner-take-all markets is a total game changer.

🤖 Nearly 60% of Americans would support the government to provide a guaranteed income to meet their basic needs, while nearly 9 in 10 want machines limited to performing tasks that are unhealthy or dangerous for humans, finds Pew Research Center in their recent Automation in Everyday Life survey. The data shows a polarisation in attitudes to core automation technologies underpinned by the neo-socialist/pro-labour rights indicated above. As usual, those with more education have had better experiences of the introduction of technology. There is a lot in this, and it certainly suggests the automation revolution will not be accepted by the broader public as eagerly as a founder will gobble down a friendly termsheet. By contrast, nearly two-thirds of Americans support human gene editing, including germline tinkering, for therapeutic use. (Also this week, a pair of British academics released a report in favour of “universal basic services” to include housing, transport, and internet access to accommodate the technology-driven disruption of the workforce at a cost of 2.3% of GDP.)

🔧 Manufacturers are making devices increasingly difficult for consumers to fix, as companies guard the gateway to the most valuable asset: the data.

The global assault on repairability highlights a bigger problem: what it means to own things in the digital age… firms now limit what people can do with the stuff they buy, in particularly the digital sort. “Owners” are often not allowed to resell it, transfer it to another devices or mash it up with other digital goods.

(And those of us with iPhones may have muttered ‘planned obsolescence’ as the device seemingly slows down around the time of the launch of the next version. Recent research suggests that there isn’t a nefarious Apple plot but rather a more straightforward explanation.)

☕ Korean instant messenger, KakaoTalk, launched an online bank that leapt to 45% market share within three months of launch. Counterpoint Research argues:

KakaoBank’s revolution can be summarized in two words, “speed” and “convenience”. Consumers had been waiting for innovation to come to the finance sector and KakaoBank has delivered this by offering speed and convenience through a smartphone. The service took advantage of its dominant messaging platform, KakaoTalk, which lowered its marketing expenses and brought a feeling of security through the familiarity of the brand.

❤️ Online dating is starting to change society. Why? Because it has the ability to connect complete strangers, irrespective of their existing cliques. Two possible outcomes suggest researchers Ortega and Hergovich are a huge increase in inter-racial marriage and potentially even stronger marriages.

😧 China wants to build a facial recognition database that can recognise _every_citizen within 3 seconds. That is 13 Terabytes of facial data, according to the documents, which works out at about 10 Kb per face. Some public lavatories in Beijing also use facial recognition so that the automatic dispensing machines will deny toilet paper to people who ask for it more than once within a given period.

DEPT OF PLATFORMS & TRUST

As larger parts of the economy move to gig or contract work, trust ranking becomes an important feature of the economy. Trust ratings have already proved their mettle on eBay, and are a key feature of Uber and Airbnb. For better or worse, they putatively allow platforms to manage (often at scale) their pools of workers.

Gavin Kelly reckons that gig workers should be able to keep their reputation ratings because:

Ratings crystallise hard-won reputations; they are the passport to future earning power. Lose them and, regardless of experience or prior standing, you are pretty much starting from scratch.

Author on peer-to-peer internet trust, Rachel Botsman, argues that algorithmic approaches using extant data can help — by providing reputation ratings to users new to a platform. How else will we deal with strangers “in a world in which we can find someone to fix a leak or drive us home or date with a few swipes of our phones, online trust is set to get faster, smarter and more pervasive.”

The road to platform neutral, portable ratings have a short-lived history. RapLeaf was perhaps the most well-known example. My own firm, PeerIndex, built a cross-platform reputation score before the policy decisions of those platforms to pursue closed garden networks rather than open Web 2.0 services. (I even gave a talk on the subject back in 2011.) These early firms struggled with the closed-garden and the comparatively small size of the peer-based digital economy.

But other hurdles remain. Do we trust third parties to maintain fair and redressable ratings? Neither the consumer credit nor bond ratings agencies haven’t shown themselves to be great exemplars.

That isn’t all. Many services are more than simply transactional. Rather, they are subject to Polanyi’s Paradox: “we know more than we can tell”. Translated to ratings this means as service providers, we can’t encapsulate our reputation or trustability in a single number, and as customers we probably can’t always express exactly what we want in a single number.

Laetitia Vitaud puts cleaning and janitorial services under this lens in a lovely essay:

The problem is that it’s very hard to industrialise cleaning. Unlike factory work on assembly lines that can be replicated and directed for maximum efficiency, cleaning doesn’t necessarily take place in identical environments. No two homes or offices are exactly the same. Applying the exact same moves to these different environments is suboptimal. The result is that Taylorist cleaning can only produce the appearance of clean rather than actual clean. The work is botched. Nothing is ever in-depth.

Laetitia’s solution is also ratings, but echoing Gavin Kelly, ones designed around the individual service provider. And, of course, one future might be to reinforce ‘trust’ with cryptographic proof as provided by blockchain technologies.

Elsewhere:

• Socnet’s “People You May Know” feature has long been a creepfest, connecting psychiatrist’s patients or unearthing unknown family ties. Kashmir Hill explores how the PYMK feature of Facebook is unwittingly unmasking sex workers.
• Airbnb brings a new definition to full-stack startup. It is now building apartments in Florida.
• Domestic robots will likely look cute in order to help us feel more comfortable and trustful with them. It’s a trap.
• The Indian state of Andhra Pradesh is another government looking at using blockchain for its land registries.

DEPT OF FUTURE MOBILITY

We’ve long argued that urban mobility is changing. The future will be increasingly heterogeneous. Trips will become increasingly multi-modal and the number of accessible modalities will increase: bike-share, buses, autonomous fixed routes, ride-sharing, trains and more.

Here is a great interview between urbanist Richard Florida and the authors of Faster, Smarter, Greener on the mix-modalities of future urban mobility:

The idea that we need to transport a 165 lb human in a 3,300 lb car seems wasteful. We think that enabling greater heterogeneity of urban transit options is one of the keys to reducing dependence on the single person automobile. When walking, biking, buses, railways, sharing services, and a wide range of lower-carbon vehicles such as micro- and mini-cars co-exist with cars, and are connected in an urban environment, mobility options and utility can be enhanced significantly.

I was fascinated to learn that 3.7% of the German population already owns a “pedelec” (or battery-assisted pedal-powered) bicycle. This represents a €10bn consumer market already, which analyst, Horace Dediu, reckons is poised for the ‘hockey stick’.

Elsewhere:

• How map-making is changing under the demands of machines & machine intelligence. The future will be “a prismatic collection of mappings, that invites comparison and appreciation of the ways in which our world is both known and unknown” says Shannen Mattern. (Lovely long read.)
• NVidia announces the 320 Teraflops, 500W self-driving GPU rig, Pegasus. While pricing isn’t announced, this is well above the 50 Tflops minimum that many in the self-driving industry believe is what is required for level five autonomy in a wide range of environments.
• Horace Dediu digs deep on the Model 3’s production woes. Model 3 woes notwithstanding, Tesla leads the back by volume and value in EV shipments in the US.
• Shell looks to move beyond petroleum by acquiring European electrical charging outfit, NewMotion. (FT paywall)
• Google’s Waymo is preparing to launch self-driving taxi service.

Full issue 135 is available here, and you’ll find all the past 134 issues right here.