Machine learning, invention & science; xenodesign; public data platforms; terror & technology; automation anxieties; graphene, allergies & lice++ #209
The silence of internet execs
Exponential View
Azeem Azhar’s Weekly Wondermissive: Future, Tech & Society
Two weeks to network launch!
Read how Ocean is enabling businesses to give value to data and reward data contributors.
You are reading the free edition of Exponential View.
We hosted a phenomenal conversation with the micromobility expert, Horace Dediu, during the members-only State of the Exponential briefing.
Key takeaways:
a) The purpose of micromobility is to offer maximum freedom of mobility in the least impactful way;
b) The addressable market for micromobility is worth more than $1.4 trillion annually in the US alone, compared to $1.1 trillion long-distance transport addressable by cars;
c) The industry has a number of challenges to overcome, including figuring out the pricing, business model, vehicle longevity (especially of scooters), weather impact, etc. With proper policy action, the transition towards lower-cost, lower-impact vehicles will ultimately contribute to safer, cleaner and healthier cities and communities.
Readers have a unique chance to access the full briefing through the notes we prepared for you and the recording of the talk in this free recap.
Our next briefing on Friday 22 March, with Reilly Brennan, Partner at Trucks VC, also touches on transportation, but looks closer at prospects for autonomy.
Following Reilly's briefing, we will welcome Dr Emily Shuckburgh, the climate scientist and mathematician based at the British Antarctic Survey; Sarah Hunter, the Head of Policy at Google X; and Stian Westlake whose book “Capitalism Without Capital” was one of Bill Gates' favourite books of 2018.
If you subscribe in the next week, before EV enters its fifth year on March 28, you can participate in these thought-provoking conversations for $90/year. After that, the price goes up to $99/year.
Dept of the near future
🌈 “One advantage: machine learning’s often quirky imagination. Maybe it will go in a different direction that a human wouldn’t go in”. How machine learning is changing invention. Also, an overview of ‘generative modelling’, a novel technique for scientific discovery between observation and simulation, and its impact on a range of scientific disciplines. “Will the future of science eventually necessarily be driven by machines that operate on a level that we can never reach?”
💯 EV reader, Geoff Mulgan, puts out the clearest overview data trusts that I have yet read. Data trusts, new institutions, to facilitate the legal sharing of data for better public outcomes. (I believe we'll see a vast growth in data trusts as new public institutions over the coming years. See also, Hetan Shah makes the case that technology companies should become data stewards rather than data owners. )
🎛️ Lovely profile of the complicated relationship between DeepMind and Google by Hal Hodson.
👾 Johanna Schmeer on xenodesign, a very interesting idea. “[H]uman-centered design, which for many years has been the predominant paradigm within the field of design, has in part contributed to contemporary environmental and social problems through its servicing role within the systems that created it.”
🐉 How China is rapidly moving from the IP theft to the innovation stage of economic development. It is now becoming a leader in AI, 5G and other key technologies.
🔥🌏 Burning planet: 412.30ppm
Each week, I’m going to remind us of our level of the CO2 in the atmosphere. We must avoid a level of 450 parts per million for a chance to keep global warming below 2°C.
This week’s level: 412.30ppm (12 months ago: 408ppm; 250 years ago, est: 250ppm; likely to exceed the target in 10-15 years.) Please share with your network on Twitter, LinkedIn or via email.
Elsewhere:
Bill Mckibben: A future without fossil fuels is inevitable, but the path won’t be easy.
How Britain cut its carbon emissions more than any other country.
Dept of GAFA
The Furman Report for the British Government recommended greater government intervention, including a “pro-competition digital markets unit”, greater scrutiny of mergers, enforcement against anti-competitive behaviour, data sharing and interoperability, but stops short of recommending a break-up of the tech giants.
Jason Furman points out that: “In December 2018, over 92% of UK online page views that originated from a search engine were estimated to come from Google... Over 73% of page views originating from a social media site came from Facebook or Instagram in December 2018.”
The Financial Times comes out broadly in support of the report calling it “a valuable contribution, with a blueprint for ensuring competition in a sector which has struggled to safeguard it” and “[o]pening access to Big Tech’s immense data pool would be of huge value to start-ups including in innovative areas such as Artificial Intelligence.”
I’m in broad agreement. In particular, I have long espoused the importance of thinking about data as more than a corporate, private asset, but rather considering open data can be a public good or part of the public infrastructure (see many previous EVs, including this, & Geoff Mulgan’s report above.)
We’re in strange territory. These firms are a key part of the public infrastructure. Our access to resources and services is mediated by the technologies of Apple and Google Android. The GAFA, and their smaller chums are often the only institutions with the capabilities of dealing with global cyber attacks—and often bear the burden of attenuating or deflecting those attacks.
This unfamiliar terrain merits novel rules, such as wider data sharing, to be enforced on the GAFAs.
Still, I do wonder if we can find novel structural remedies that treat internet platforms as utilities, upon which an ecosystem of competitors can thrive. The recommendations in the Furman report that platforms open their data to competitors are a step towards that. For a good precedent of this look to Open Banking in the UK. This forces banks to offer consistent API access to consumers’ accounts, which in turn is creating a novel suite of finance apps, focused on consumer needs. Another example in the UK has been the regulatory unbundling of our phone system into a utility network and a separate customer-facing services company, which appears to be working.
Separately, EV reader, Rowland Manthorpe, makes a good case that the large tech platforms excitable practice of buying cooler companies is a problem. In a decade, the five biggest have acquired 400 startups, Google at a rate of 1 every 18 days.
📦 Zack Kanter: What is Amazon? (Long read.)
So, what is Amazon? It started as an unbound Walmart, an algorithm for running an unbound search for global optima in the world of physical products. It became a platform for adapting that algorithm to any opportunity for customer-centric value creation that it encountered. If it devises a way to keep its incentive structures intact as it exposes itself through its ever-expanding external interfaces, it – or its various split-off subsidiaries – will dominate the economy for a generation. And if not, it’ll be just another company that seemed unstoppable until it wasn’t.
EV reader John Battelle reckons that Facebook’s move to privacy is designed to capture “the entire internet”.
By declaring “private conversations” as its new business model, Facebook can undermine the app store model driving all of mobile, unseat Amazon as the king of e-commerce, hollow out Google’s control of search, nip Apple’s transition to services in the bud, take a vig on every transaction across its ecosystem, and insinuate itself into the private, commercial, and public lives of every citizen on the Internet.
Elsewhere:
Tumblr only lost 17% of its traffic after it banned porn.
Twitter is testing a new app designed to produce “healthier conversations.” 🧐
Internet content providers, like Microsoft, Google & Amazon, now lease about half of all subsea bandwidth. (Fab essay.)
Dept of autonomous systems
Early evidence suggests the sad demise of an Ethiopian Air flight this week is of similar pathology to that of Lion Air 610 in October 2018.
It was a Boeing 737Max which crashed in both cases. This plane can be aerodynamically unstable and needs an automated system called the MCAS to fly. The MCAS system was implicated in the earlier Lion Air crash. Without MCAS operating correctly, the planes can crash regardless of the pilots’ efforts. It becomes machine over man. The planes are now grounded and Boeing is rushing out a software patch, as this excellent report explains.
Autonomous vehicle makers are struggling with their transition to fully autonomous vehicles, (see my 2018 predictions on this). It turns out that the transition from Level 2 autonomy, which is really enhanced human assistance, to Level 4 autonomy, geo-fenced bull autonomy, isn’t so much level 3 autonomy but a chasm, in which lies mortal risks to humans.
A number of Tesla drivers have now died with autopilot activated, and while many autonomous vehicles teams have abandoned Tesla’s approach, Elon Musk’s firm is sticking to its previous strategy. Timothy Lee argues that this is problematic as “Tesla’s plan is to essentially run a massive driverless-car testing project using its customers as unpaid safety drivers. Drivers get no real training on the dangers of inattentive Autopilot use.”
The problem with level 3 systems is that they work, without human intervention, until they don’t work. And when the machine doesn’t work, the human needs to step in. And the human may not have time to step in. I actually wonder whether level 3 autonomy, conditional autonomy, where a driver needs to be ready to take over at any time, can exist in any meaningful context. I find it hard to understand how insurers could insure such a vehicle, and how regulators could allow them to exist. You can read a more detail analysis of these points in this excellent review here.
Writing in Harvard Business Review, Nicholas Oliver and his collaborators suggest that “automation can limit [humans]’ abilities to respond to such incidents, as becoming more dependent on technology can erode basic cognitive skills.” Acculturated to the autonomous system doing their work, that they are unprepared to take over. (This seems to have been the case with Air France 447, which crashed in the South Atlantic a decade ago.)
And one last, and meaningful hurdle, autonomous vehicle systems will need to overcome is that system level testing argues Paul Orlando to avoid “chaos at scale”.
An academic study showed that fast-working robots could demoralise workers.
Short morsels dropped on the floor to improve your immune system
🇪🇪🌟 In Estonian elections, 44% of votes were cast electronically.
Culture as a form of collective intelligence.
🦴 Repairing broken bones with graphene.
🔥 The extensive web of misogyny. Do read.
💎 Emma Haruka Iwao, a cloud developer advocate at Google, calculated Pi to 31,415,926,535,897 digits. 💎
“I tell people, when they drop food on the floor, please pick it up and eat it”: on the rise of allergies. 😋
Feather-eating lice quickly evolve to match the colour of the bird they're hanging out on.
Emmental raised on hiphop tastes better than emmental raised on Mozart.
👌 This conversation between Yascha Mounk and Karen Stenner on the connection between authoritarian personalities and populism is exceptional. It provides a helpful framework for understanding the global rise of populism under the pressures of technology-mediated change & growth of social media services. Listen with critical faculties engaged.
End note
I usually write Exponential View on a Friday afternoon. This week I was shocked by the terrorist attack in New Zealand. My condolences and prayers to those affected.
We’ve clearly got a difficult road ahead of us. This sober first take by CJ. Werleman on the “broken white men and the racist media that fuels their terrorism” is worth reading.
But I want to turn my attention to the digital platforms, YouTube and Facebook in particular, and their relationship to the spreading and distribution of extremist content and, in particular, to the video of the murders. And please notice the reality of the choices made by those platforms in the context of the pace of technological enhancement.
It was three years ago that Facebook had to deal with the live-streaming of the aftermath of the shooting of Philando Castile. This is the same Facebook that regularly took down photos of women breastfeeding their babies. Or that rapidly censored the world’s most famous war photograph, because it showed a naked girl burning under napalm. (I wrote about this in EV#78). Three years is a long time for tech firms, especially with the breakneck pace of technological development and for companies which pride themselves in their speed of innovation and execution.
There are very few circumstances that the unfiltered video of the massacre should have been widely distributed the way it was. Yet it was. BBC reporter Jim Reed found, YouTube took more 75 minutes to remove a copy of the video, after he flagged it to them. It was watched 20,000 times in that period.
This balanced review in Wired explains why the scale of these Internet platforms, amongst other things, present an Augean task to clean the feculence of such social sharing. Alex Stamos, who used to run Facebook’s security systems, nuances this—in some sense these videos got wide distribution because people were actively searching for it, goaded on by the TV networks.
Yet, YouTube has been contending with extremist content of all types for several years. As Ben Collins points out, both Facebook and YouTube have been exceptionally good at stopping certain types of extremist material. And researchers have for years been warning online platforms of that “online extremism and radicalization results in real-world violence.”
Try posting a fake advert masquerading as a major global brand on an Internet platform like Facebook or Twitter. You won’t succeed—because adverts, even those worth a few hundreds of dollars—go through a review process. This review process can sometimes take hours.
At the same time, for all the claims that it is technically hard to identify and automatically block this content, remember that we are seven years into the deep learning boom. Google (which owns YouTube) has already built an AI system that defeated humans at Go, about a decade earlier than most computer scientists believed possible. Facebook analyses billions of images per day using machine vision, and is now building its own AI chips.
And then consider the intentions of the people who designed these things. These products seem to be operating as designed. As Mark Zuckerberg said when he launched Facebook Live, “we built this […] so we can […] support whatever the most personal and emotional and raw and visceral ways people want to communicate” (my emphasis). A veteran media commentator, Mathew Ingram, said: “Literally everyone who has thought about this kind of thing for more than five minutes knew when Facebook launched FB Live that it would eventually be used for this.” And ultimately, these products could be designed differently to support different goals. They could be build in such a way as to carry their full costs internally, rather than to externalise them in myriad unpleasant ways.
More than a decade ago, well-heeled banks built toxic financial instruments, designed by highly-talented, well-paid employees. These promised amazing returns. Then they exploded and left society to clean up the mess. The spillovers from today’s internet platforms echo that recent past.
Azeem
This week’s Exponential View has been supported by our partner: Ocean Protocol.
Ocean Protocol is kickstarting a data economy that is global and borderless.
Don't miss out! Be part of a New Data Economy today.
What you are up to—notes from EV readers
My mum turned 79 this week. She is raising money for the UK’s Mental Health Foundation. You can donate here.
Dan Gilmor on recent experiments that show how social media platforms can help undo the sharing of false information: “The platforms have everything they need to help corrections catch up with mistakes, and it would be to everyone’s benefit if they’d deploy the tools to make it happen”.
Horace Dediu's Micromobility project is on Kickstarter, and so close to making the goal!
Nicolas Colin reviews 10 books on entrepreneurship, finance, state, democracy & more, including the works of Mariana Mazzucato, Bill Janeway, and Hilary Cottam.
Rhea Mehta, CEO of Bowhead Health announces the launching of an encrypted DNA test that lets customers own, manage and monetize their own genetic data using blockchain.
Stephanie Greg and Marios Yer launched a non-profit project, Health 101, to get you regular reminders for healthcare screenings based on your age and sex.
Rachael Blair on what digital agencies can learn about business ethics from the Fyre Festival.
Diana Wu David published a book aimed at 55+ population grappling with the future of work. The book prepares people for growth mindset and practices for success in a fast-changing world.
Nick Patience and his team at 451 Research recently published reports on top use cases for AI and ML in key industries and on AI and ML enterprise adoption strategies.
Andrew Burgess: What would more democratic AI look like?
Send your news and projects to marija@exponentialview.co