🔮 Breaking up Facebook; growing markets; GDP & control; ignorance; lasers, swords & ‘shrooms++ #217

Agnotology and social media

Exponential View

Azeem Azhar’s Weekly Wondermissive: Future, Tech & Society 

This issue has been supported by our partner: Ocean Protocol.
Ocean is supporting the development of pan-European data sharing space.
Read more here


You are reading the free edition of Exponential View

If you enjoy the newsletter, you can become a Member where you will have access to our live, interactive briefings with world experts (such as Bill Gross, Rumman Chowdhury or Horace Dediu); as well as bespoke insight and analyses (and many more goodies).

You can become a member for $100 / £77 per year, or $12 / £9 per month.

We have special discounts for academics and students—for more information, email support@exponentialview.co.

If you don’t want to enjoy the benefits of membership—which are awesome and worth much more than the sticker price—I would still like your assistance in helping Exponential View flourish. Actively contribute to the success of this endeavour by helping it grow. You do this by regularly emailing a recommendation to several friends. Please do this, and don’t rely on others to. Thanks!

EXPERIMENT: Since we have covered many of the topics we explore each week in our podcasts and briefings, we’ll link to them where there is a relevant conversation. Look for the 🎧 icon for podcasts and the 🎓 icon for an expert briefing.


Dept of the near future

👍 Facebook co-founder, Chris Hughes, says with great clarity what so many now believe: Facebook should be broken up.

Facebook’s dominance is not an accident of history. The company’s strategy was to beat every competitor in plain view, and regulators and the government tacitly — and at times explicitly — approved.

I don’t blame Mark for his quest for domination. He has demonstrated nothing more nefarious than the virtuous hustle of a talented entrepreneur. Yet he has created a leviathan that crowds out entrepreneurship and restricts consumer choice. It’s on our government to ensure that we never lose the magic of the invisible hand. How did we allow this to happen?

This is more than simply restricting consumer choice. Facebook’s quest for growth has also spread Zuckerberg’s dorm room cultural values across the globe. It has become an interface between people as citizens (not merely as “consumers”) and the resources they need to access. Facebook has also been instrumental in the growth of agnotology as a business and societal disease.

🕳️ danah boyd on how social media fosters agnotology, the 'strategic and purposeful production of ignorance' and is a 'tool of oppression by the powerful.

What’s at stake right now is not simply about hate speech vs. free speech or the role of state-sponsored bots in political activity. It’s much more basic. It’s about purposefully and intentionally seeding doubt to fragment society. To fragment epistemologies. This is a tactic that was well-honed by propagandists.

One nuance I would add is that while actors within these platforms may act to purposefully spread ignorance, I think the platforms themselves have apathetic positions on epistemologies. Rather this emerges as a result of chasing engagement and the ad-supported business model. (We first covered agnotology in EV#24.)

📈 Entrepreneur Elad Gil: Markets are bigger than ever before and even “startup companies will continue to grow in scale, and get there even faster.” He reckons that software markets are 10x bigger now than 10 years ago, and getting bigger. 🎧 Elad and I discussed growing at scale a few months ago. (Also, incredibly, the Softbank Vision fund is up 45% in two years on a book-value (not realised gains) basis. Impressive for a fund this size.)

🤔 Related, Casado and Lauten argue that data moats may not extend to a large class of companies and that in many cases data network effects, much lauded for creating impregnable businesses like Facebook and Google are simple scale effects. It’s an interesting argument with some merit. I think it misses at least one trick: incremental data can provide for better segmentation and, thus, novel products or pricing targeted at specific niches.

💯 Is time up for capitalism? Filmmaker, Astra Taylor, is typically thought-provoking:

[W]e’re up against ecological limits, not monetary shortages; we are constrained by a carbon budget not a federal one, and we need to remake our economy to reflect this reality.

🎧 Economist Kate Raworth discussed similar ideas back in November.

🔮 MUST LISTEN Steve Hsu and I discuss the intersection of genetic engineering and machine learning. Hsu is a pioneer in the technique of polygenic risk scoring, which determines disease predisposition based on a number of genes. It is not a new technique, but it’s been accelerated by cost-declines in gene sequencing and advances in machine learning. As this technique is making its way into IVF clinics where researchers can determine traits such as intelligence and looks in embryos, we need to ask: where is the boundary between the desire to enhance wellbeing and the ambition to create a flawless human being?


This issue has also been supported by Wired 💥

Save the date for Wired’s exclusive one-day event about computer creativity and consciousness, ethics for algorithms, and autonomous decision-making.

EV readers get exclusive 10% off—get your tickets here before it’s too late!


⚠️🌏 Burning planet

Each week, we’re going to remind you of the CO2 levels in the atmosphere. We must avoid a level of 450 parts per million for a chance to keep global warming below 2°C. If we don’t change how we do things, we’re likely to exceed the target in 10-15 years.

  • The latest measurement (as of May 9): 414.46 ppm; 12 months ago: 409ppm; 50 years ago: 326.66ppm; 250 years ago, est: 250ppm.

Share this with your friends by forwarding this email or via Twitter.

🕛 The UN Environment Finance Initiative reckons that delayed decarbonisation will cost the world’s 30,000 listed companies $1.2 trillion in higher costs.

In addition, the IMF calculates that fossil fuels are subsidised to the turn of $4.7 trillion globally, representing 6.3% of GDP. Efficient pricing would reduce global carbon emissions by 28% and increase government revenues by 3.8%

The journey to zero carbon is fraught with geopolitical risk, with many losers on the way. Good scenario-based commentary by Bazilian, et al.

Dept of GDP

EV reader, Frances Cairncross, pointed me to this excellent speech by the Bank of England’s Andy Haldane, on whether all economics is local, and critiquing the limitations of aggregate GDP as a measure of economic health.

The limitations for GDP, which emerged as a standard tool for measuring the health of economies at Bretton Woods, have been long discussed. We've covered them in this newsletter. See, for example, Diane Coyle’s excellent guest edition of this newsletter or my discussion with 🎧 Marianna Mazzucato.

Haldane makes some important realisations. Our economies are complex adaptive systems in which

behaviour  [...] is difficult to predict ex-ante, especially at times of policy change; it is emergent, just as a hurricane or tornado is emergent. Often, our policy intuition about complex systems is simply wrong. No model, however micro-founded or data-rich, is proof against those uncertainties. But one that embodies complex, micro-level dynamics is more likely to do so than one without them. A complex systems framework can make for robust policy choices.

Our economies, like our politics, are local. Like the seashore, the more you magnify an economy, the greater its richness, complexity, self-similarity. Like our bodies, understanding our economic health means taking readings at many resolutions. It means understanding the moving body parts, and their interactions, in microscopic detail. It calls for new data, at a higher frequency and higher resolution, and new ways of stitching it together. It means making micro-to-macro a reality.


Haldane goes on to conclude:

our economic policies would be better able to serve the public, and better understood by them, if we could do [model the economy in microscopic detail] in close to real time.

I reckon we’ll be able to do this better going forward because we’re rapidly creating digital twins of the real economy. We have increasingly global visibility of every package or manufacture of the global economy. Digital services like Salesforce, Xero or Stripe are capturing real-time invoicing, payment and settlement information of our businesses. Google searches allow us to build predictors of likely behaviours, albeit stochastically rather than deterministically represented.

In fact, modern machine learning, optimisation and agent-based modelling approaches now have the data and the computational power to run increasingly large-scale models with more and more agents. Firms in the UK, like Prowler.io (terrible name, decent science) and Deepmind, are actively researching and building simulation systems in the intersection of reinforcement learning, agent-based modelling and probabilistic systems. Unity, which this week completed a funding round valuing the company at $6bn, builds a widely-adopted toolkit for some classes of this type of simulation. The scale of what modern machine learning can do to model these systems is quite something. The global economy could soon be in reach.

Previous attempts to model the whole economy—think of the Soviet’s Gosplan—failed, in part, because you couldn’t gather accurate data in a timely fashion and, even if you could, you couldn’t solve the optimisation problem of that data. (They failed for many other reasons, too.)

But even if we can model this complex adaptive system of the economy, we’ll need to be careful. In The Road to Serfdom, Friedrich Hayek, writing before the establishment of complexity theory and our ability to model that complexity in silico, believed it would be “impossible for any mind to comprehend the infinite variety of different needs of different people which compete for the availability of resources and to attach a definite weight to each.”

He goes on to describes the process by which the allure of economic planning, and the centralization of power, runs counter to liberal ideals and the needs of a democracy. Planners turn into nudgers. Nudgers nudge harder. And what holds the nudges in check from authoritarianism and then totalitarianism?

Of course, we’re quite far from that slippery slope. A better understanding of the complex adaptive processes within our economy out of which structure emerges will help policy makers, and hopefully our politics too. It might even achieve what Andy Haldane hopes: create economies which are “better able to serve the public.”

But Hayek’s analysis often springs to mind when I get to grips with progress in complexity economics and our growing ability to simulate it.

If you are interested in this topic (I hope you are, I just spent 650 words on it!), I am running a session on the intersection of complexity theory, economics and artificial intelligence at CogX in London on June 12. And early next month, the podcast will have a discussion on what comes next for complexity economics with Brian Arthur, one of the earliest proponents of the field.

On the subject of rethinking GDP for the digital age, I missed writing about Erik Brynjolfsson and collaborators work on GDP-B. It is a mechanism to capture the value of things that we don’t pay for, like online maps, Wikipedia or search results. The problem is that GDP measures what we spend, and Brynjolfsson wants to measure the Benefits we gain, hence GDP-B.

One result arises from a set of experiments providing estimates for how Dutch students value various products per month: €536 for WhatsApp, €97 for Facebook, or €1.52 for LinkedIn. Americans were cheaper, wanting compensation of $42.17 per month or $507 per year for Facebook. If this value was captured in traditional GDP it might have added an additional 0.11% to US GDP, reckon the authors. I question the process of assuming the experimental subjects could accurately gauge their own preferences; in general we’ve found people are not consistent, and poorly understand their own (inconsistent) utility functions. I recommend reading Gillian Tett’s analysis of the Brynjolfsson, and other approaches, from 2018.

While I think GDP-B is an intriguing avenue of exploration, I remain unconvinced that finding better ways to dollar-measure economies that are already mind-blowingly productive. This conception of GDP still suggests that everything that matters can be measured against a dollar yardstick. I also wonder whether GDP or GDP-B, this sort of ambient temperature of economies that are actually made up of hot and cold vortices and eddies, interacting and discombobulating, is useful in its conception at all. (See the excellent argument made by Andy Haldane at the top of this section.)

My take is that the real work has to be somewhere closer to understanding how to move the economy to better meet human needs within the environmental boundaries, and how to find the correct balance of market and non-market institutions to do that.


Dept of tech nibbles

Geoff Hinton, the deep learning pioneer, speaking at Google’s IO: “We humans are neural nets. What we can do, machines can do.”

✨ A good introduction to Judea Pearl’s work on causal reasoning and why we might want to introduce causality into our AI systems by Alexander Lavin.

🚨 Overview of how AI is being used to combat human trafficking and exploitation: “They can do in seconds what could take investigators months or even years.”

I recently discussed ethics and AI with Joanna Bryson. Conversation notes and transcript are now available to all Premium members.

GM’s Cruise raises $1.15b at a $19b valuation from Softbank. The prize for winning the autonomous transportation market is going to be huge. And remarkable that even a large firm like GM needs to find external capital to fund their ambition. Transportation investor, 🎓 Reilly Brennan, told EV members that he reckons “it's going to take about 20 years and about $40 billion to get to fully autonomous vehicles” for each successful car platform.

💓 A robotic surgical device has learnt to autonomously navigate inside a beating heart.

Facial recognition technology used by London’s Metropolitan Police incorrectly identified members of the public in 96 percent of matches made between 2016 and 2018. (Meanwhile, US airports are on track to use face recognition on 97 percent of all passengers within four years.)

Overview of Nextdoor and other neighbourhood social media apps which are based on socialising around fears of crime…. as rates of crime in the US decline. It will be curious to see if the spread of these apps will exacerbate marginalisation, racial stereotyping and unfair policing. (Narrator: “They will.”)

Global smartwatch sales are up 48%.


Short morsels to appear smart at dinner parties

Vietnam, China’s closest ideological neighbour, is rolling out its own 5G network using some homegrown technology, some foreign technology and no Huawei.

🤳 All those warnings about screen time before bed being bad for teenagers might have been wrong after all. The same study also busted the myth that social media is responsible for making teens unhappy.

🥜 Eating nuts during pregnancy could boost your child's intelligence.

Power laws abound. The top 1% of performers took 60% of all concert ticket revenue globally in 2017.

By 2021 there could be a 3.5 million shortfall in the number of cybersecurity workers needed globally.

🍄 By a margin of 0.6%, Denver has voted to decriminalise the use of magic mushrooms.

😷 Beijing-based Bitmain appears to be significantly scaling back its bitcoin mining, generating a whopping 88% less computing power than a month ago. A sign of rationalisation in the market?

The US is a step closer to being able to put anti-missile laser weapons on its fighter jets. (They have also put extrusible ninja-swords on drone-launched Hellfire missiles to kill targets and reduce civilian casualties.)

〰️ Researchers have proved that antimatter is both a particle and wave, in keeping with quantum physics theory.


End note

This one is especially for the 10,000 or so new readers to Exponential View this year, to help explain what I’m doing here.

This newsletter is an artisanal product. What I’m trying to do is improve my understanding of the future by looking at the key questions from different angles.

There are commonalities each week in the newsletter. But also differences. Like an artisanal restaurant, whose menu varies by the produce available, EV varies each week based on what I come across, how much time I have and how I feel. You will get a flavour for it over two months, not less.

I encourage you to read it slowly, rather than skim it fast. Give yourself time. You’ll find connections across items (and across weeks.) consider the connection between the report on crime-obsessed social networks and danah boyd’s argument about agnotology? Or the contrast in views on AI between Geoff Hinton and the causality school? Or Astra Taylor on capitalism and Eric Brynjolfsson on GDP-B? What do those tell us?

There aren’t really right answers, but there are good questions :)

Cheers!
Azeem 😎

P.S. Below are some amazing news from fellow EV readers. Do check them out!

This issue has been supported by our partner: Ocean Protocol.
Dubai is creating a decentralized and
truly open data infrastructure and Ocean can play an integral part.
Read more about smart city infrastructure in the new report by Outlier Ventures & SmartDubai.

What you are up to—notes from EV readers

Congrats to Valentina Milanova for closing a $5.5m funding round for her (quite remarkable) pain-relieving tampon startup.

David Ticolli’s report on the future of work proposes a comprehensive systems framework for modelling the future of work.

Congrats to Emily Shuckburgh for leading Cambridge University’s new Centre for Climate Repair.

Chris Wigley of McKinsey’s QuantumBlack discusses purposeful business practices and ethics for AI on this podcast.

Oz Woloshyn shared a podcast he’s creating: Sleepwalkers explores the thrill of the AI revolution hands-on.

Georgia Ritter has launched a community for people building side projects. Join!

Mike Walsh’s book “The Algorithmic Leader: How to Be Smart When Machines Are Smarter Than You” is out now!

Kris Oestergard’s book “Transforming Legacy Organizations” is available for pre-order now.

Arif Khan’s commentary on Chris Hughes call for Facebook breakup: “what the article misses, however, is a forward-looking analysis on Zuck's desire to dominate the next battleground: our money supply.”

EV readers Zavain Dar and Evan Nisselson discuss the future of computer vision.

Wayan Vota on the must-ask questions when developing AI.

Want to share your project or news with the EV community? Email marija@exponentialview.co