Copy
#169 Principles of AI | FWD to a friend. 

Exponential View

Azeem Azhar's Wondermissive: Future, Tech & Society
  This issue has been supported by our partner:
Kindred Capital
Dept of the near future
⚠️ Microsoft has acquired Github, a service which allowed software developers to collaborate on code, for $7.5bn. Github has become an essential part of the global development community and is the space where much of the collaborative work in the discipline takes place. The value for Microsoft is in owning the platform where opensource developers around the world congregate, as Paul Weinstein argues. Microsoft has long understood the value of developers but the GitHub acquisition is a major departure. (This is a major megatrend touching on platformisation, the growth of ecosystems and the blurring of the Coasian boundary of the firm. Frankly, Github will only survive this acquisition if it remains credible to the community, which means Microsoft leaving very well alone and gently working out how to create, rather than extract, value.)

👛 Capital’s share of income in the US is higher than we think. About half of US household income is derived from wealth. And 60% of that wealth is inherited. As Steve Roth says, this is "pretty hard to square [this] reality with the widely held view of prosperous, hard-working makers vs indolent do-nothing takers. Who's who?"

📚 Automation will make lifelong learning a necessary part of work.

🧐 Google announced its AI Principles which include being socially beneficial, avoiding unfair bias, safe, accountable, transparent, scientifically rigorous and not used harmfully (including for weapons or surveillance beyond ‘international norms’). It is a good start, recommend you also read their more detailed guide to Responsible AI Practices. Importantly, Google doesn't explain what the governance system around these principles will be. Who will decide what counts as harm? Or what as beneficial? And will there be remedies? I’m also intrigued by how this describes a more complex set of objectives for the firm than the definition so much of recent American capitalism has hung its hat on that of Milton Friedman’s claims about the social purpose of the firm. (These are also only Google’s AI principles, so I wonder whether Alphabet, the parent co, may spin up AI companies which support invasive surveillance or military applications.)

🔥 An inside look at the campaign by Google employees to cut a controversial Pentagon contract. Super relevant for how politics may increasingly get done amongst these monopo-platforms with their highly trained, well paid and, often, politically alert workforce. And since these mega-platforms can rival nation-states in some domains, this really does become politics governed by a strange kind of epistocracy within this founder-driven fiefdom. I know Facebook employees have sometimes stood up to Zuckerberg. I'm curious about the culture in Apple and Amazon & the BATs and the extent to which these sorts of initiatives might take hold there. Do let me know, if you have a solid perspective. 

Behind the messy WhatsApp/Facebook divorce: the founders left $1.5bn on the table. Ostensibly for strategic differences, but those differences were really about the promise made to WhatsApp users about privacy and their data and Facebook's views of advertising. The nub: this was a disagreement about ethics. 

💸 Ant Financial, the payments spin-out from Alibaba, raised $14bn, in a staggeringly large private financing. What does it mean? Investors are betting on a future configured very differently to the recent past, they are comfortable with Ant’s culture of execution, and there is far too much cash lying around (and too many investment opportunities are low yielding.) I’d argue one more thing: you are probably better off backing a market-leading segment upending fintech, with all its upstart risks, than a stodgy bank which can’t get its internal IT systems, culture or strategy together.


👾 Viability #4

Awesome support for the viability campaign. Many ask why? Well, a larger audience creates a flywheel which allows us to deliver more interesting things like a better insight, a consistent podcast, proprietary research, new analyses, community activities & tasty surprises.

We *reckon* that between 7 and 9% of our readers have helped us with referrals in the past three weeks. If you aren’t one of them, please take a moment to help. Email referrals work best: to a group of friends, your alumni network, your direct reports, your students, your workgroup. But if you are a blogger, a blog post can work. (Here’s another good example from Jana Moldovan.) Finally, shares on LinkedIn and Twitter work a treat. The URL to include in your email referrals is http://exvw.co.


Dept of firms, platforms and ecosystems
Back in 1937, economist Ronald Coase tried to work out why firms existed. Why didn't people just organise themselves as networks of freelancers? He argued that in many cases, especially for uncertain or frequent tasks it was easy to organise and direct people who were employed than to go out to the market to negotiate a fresh contract with a supplier. The Coasian boundary is that marginal point where it's cheaper to go to the market than to operate with the fiat power you have over employees.

But three trends have started to blur the definitions of "in-firm" and "out-to-market", and a new pair of structures seems to be winning support. The first is the ecosystem and partnership model where risk is broadly shared across a complex web of partners in a supply network (network rather than chain because chain feels too unidimensional to me). This is common amongst the engineering industry (such as the mobile phones or airplanes) and increasingly common in the software industry.

Firms adopt platform strategies to cajole third-party developers to build on 'their platforms' in order to drive value from their platforms. This is what is behind Microsoft's acquisition of Github, or Apple's extensive developer support. This ecosystem or open innovation thinking has become de riguer in the tech industry for more than 20 years. It'll spread to the other industry segments which survive the next two decades.

The other trend is the marketplace platform which explicitly coordinates two or more sides to provide a service. The fastest growth is in the gig economy, and the relationship between an Uber driver and Uber is not quite like that of an employee and an employer in a Coase-like firm. But it isn't like a fresh contract in a perfect market. And there in lie many uncertainties.

The third trend is that "token economies" or "token networks" designed to incentivise participants in a particular economic space, and then facilitate coordination amongst them without a traditional corporation at the centre. These blockchain-based, post-firm networks are in their early days with lots of smart people & good theory but little hard evidence at which to point.


EV reader, Sangeet Paul Choudary, has looked extensively at the digital labour platforms for the International Labour Organisation, specifically exploring how design principles and standard execution procedures might affect worker well-being. Marija Gavrilov summarised the report here but the full document is here.

It’s Sangeet Paul’s typical modus operandi: highly-structured analysis which pulls out some very interesting points:

  • The lean-startup methodology of fast, frequent changes to optimise an experience can put workers on the receiving end. (For example, workers may invest heavily based on some previous feature which is removed by the platform.)
     
  • Metric-based management, which is the hallmark of all high-performing internet firms, can create a very high watermark for continued platform access. I sometimes wonder whether it is tougher to maintain the higher Deliveroo or Uber worker rating than it is to consistently work for an employer.
     
  • Final recommendations: A regulatory framework for platforms would best be structured as multi-sided coordination between three stakeholders: the platform, the regulator and the workers no effective regulatory framework for platforms can exclude consideration of the data generated on them. 

Another pair of EV readers, Mark Bunting and Brian Williamson, have tried to understand why regulation struggles to get to grips with online markets. They suggest that platforms now make and enforce market rules, in their perception of the interests of the whole ecosystem. The platforms are often better at it than the regulators, but where they’re not, new approaches to regulation are needed. One example: transparency is good, but too much transparency might lead to behaviours that game the governance of the platform & its ecosystem. (Full report, which I’ve not read, is here.)

Elsewhere in platforms and ecosystems
  • Apple has paid out $100bn to developers via the App Store. Every week, 1 in 15 humans uses it. Apple has used this ecosystem and the community of developers to drive uptake of Swift, its programming language. It is worth understanding this App store ecosystem in the context of the Microsoft Github acquisition above.
  • Amazon has acquired rights to stream 20 British Premier League games in the UK. Access to sports rights is one of the few scarce “sure bets” these days, and it was locking up those rights that allowed Rupert Murdoch’s Sky to achieve a dominant position in the British pay-TV market. Amazon will stream these free to Prime subscribers.
  • Cryptokitties, the token for collectable digital kittens, declined below 200 users per day, just a few weeks after raising a $12m funding round. 

Short morsels to appear smart at dinner parties

💿 Spotify is starting to compete with labels by signing acts directly. Makes sense given their data & distribution advantages, I wonder if it parallels Netflix’s move from the simple aggregator to the creative principal.

💉 “She has sociopathic tendencies” Nick Bilton on founder cults and Elizabeth Holmes of Theranos.

"Culture is not a mere afterthought or an add-on, but a key contributor to urban economic growth." Super research, lovely visuals.


Using a machine learning model processing 1.5 billion tweets by 5 million users, researchers estimate personality traits and an entrepreneurial personality profile for 1,772 U.S. counties.

🌀 AI is forcing generalist chips to yield to new, specialist processors, says The Economist. (For a more detailed view of this, read my essay from last year on the subject.)

🚗💨 Health damage associated with diesel vehicles is estimated to be twenty times more than that of electric vehicles and five times more than petrol vehicles.

Investments in fossil fuels could have multi-trillion dollar losses, regardless of new emissions limits says a new report. 


The huge oversupply in photovoltaic modules should result in price declines of 34% this year.

We’re going to see cars at level 3 autonomy very soon. Level 5 may take much longer.

🐝  Bees understand zero. (We didn’t start to use zero until about 1770 BC.).

Superheavy elements and the future of the periodic table.

Black skinheads & the origins of skinhead culture.
Azeem's end note 

On Monday and Tuesday, I’ll be at the CogX festival of AI in London where Exponential View is curating one of the main content streams, The Cutting Edge. Here we'll cover the most interesting future paths for AI & its applications. We cover everything from the future of chip architectures to design principles for centaurs, from swarm dynamics to brain computing interfaces, and a lot in between. If you are around, please swing round to our stage (you can register using my special code 50EV30192 for 50% off on 2-day passes).

Need another AI hit? Check out what my buddy, Rumman Chowdhury, has been up to. She is launching a fairness assessment tool for AI systems. Should be interesting as it tackles data & model repairs to ensure sensitive variables are not wrongly emphasised.

Finally, a reminder that I’m investing actively in superb founders aligned with the many theses with Exponential View. I’m interested in Seed & Series A investments. My 18 predictions for 2018 give a rough, non-exhaustive outline of themes of note. Drop me a mail, ideally via a warm introduction from someone I trust.

Hasta la vista,
Azeem 😎 
This week's Exponential View has been supported by our partner, Kindred Capital: 
 
Copyright © 2018 ExponentialView, All rights reserved.

Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list
Having trouble seeing this email in your inbox? View it in your browser.
For commercial enquiries, please email Marija Gavrilov.