🔥🔥 Disrupting cars; where AI is heading; automating & inequality; the cause of Alzheimer's; fusion & folding phones++ #202
I discuss the open internet & democracy with Marietje Schaake
|Jan 27||Public post|| 8|
Azeem Azhar’s Weekly Wondermissive: Future, Tech & Society
This issue has been supported by our partner: Ocean Protocol.
A Decentralized Data Exchange Protocol to Unlock Data for AI
Want to know more about Ocean? Subscribe here to stay updated
Dept of podcasts 🎧
The Exponential View podcast is supported by Spotify.
My guest this week on the Exponential View podcast is the Dutch politician and the Member of the European Parliament, Marietje Schaake.
Marietje is a strong proponent of regulating the cyberspace in order to secure the open internet. We discuss this and more in the latest episode, which you can get on:
Spotify (recommended) | iTunes | Soundcloud | Stitcher | Overcast | Breaker
(Please take a moment to review the show!)
Dept of the near future
💯 Horace Dediu: Understanding the disruptive potential of micromobility Exceptionally tight analysis on how micromobility will unbundle the automotive industry. Car makers are starting to abandon the lower-end markets, essentially smaller cars for city trips, giving micromobility players a beachhead at this lower end. Horace estimates we might be 10-20 years from half of all US personal vehicle trips being via micromobility: “[b]ike sharing has been growing as fast or even faster than social media and when you look at the rate of growth of the scooter sharing.” (See also: some operational data on vandalism and damage rates for two of San Francisco scooter networks. And this, one consequence of the rise of Uber & Lyft has been the decline in the use of public transit in the US.)
🧨 Technosolutionism is not going to resolve society's ills—we need to fix capitalism instead, argues Douglas Rushkoff in this must-read essay.
⛰️ An accessible & comprehensive analysis of 16,625 AI papers suggests that research in deep learning has peaked, and researchers are moving onto other techniques like reinforcement learning. (Of course, deep learning will continue to proliferate amongst developers as a tool of choice for the next few years. And the more sophisticated will be operationalising techniques like GANs and reinforcement learning.)
🛰️ The explosion in low-cost satellites is changing the game of business intelligence. Earth's orbit now peppered with multi-spectral CubeSats gives us new levels of insights. (Azeem's comment: I've been tracking this for about three years, and the growth of private satellites is on a familiar upward trajectory, while the price for accurate near-real-time monitoring is declining rapidly.)
🌐 McKinsey & Co put out a detailed brief on the state of globalization. Centre of gravity is moving South and East; as physical trade retrenches, data flows increase; technology disruption accelerates, creating more opportunities which should get captured & shared. (Azeem's comment: I think this is worth reading critically as some of the analysis is sound. The conclusions are not though: it is predicated on the model that growth trumps everything; sees the tribalism purely through the lens of inequality and has little to say about climate change. Ultimately, it all feels a bit "faster horse-ish" rather than "new paradigm.")
Dept of artificial intelligence and data
DeepMind announced another seeming-breakthrough in applied machine learning when it released AlphaStar, an AI system, that could beat a world-class human in the game of Starcraft II. This is quite an impressive research engineering milestone. Previous efforts, AlphaGo most notably, won in a game of perfect knowledge (that is, all the available information about the game was available to all players). Starcraft is a more complex game. It involves partial information, so players need to scout to understand their environment. Players face a cornucopia of choice at any time. They play live against dynamic opponents. And successful strategies often include choices with payoffs far into the future. The state-space complexity of Starcraft II is 10^1685 or about 1000-orders of magnitude more complex than Go; which in turn was about 100-orders of magnitude more complex than Chess. It took us 18 years to advance from mastering Chess to mastering Go. It has taken us four years to master Starcraft II. What this tells us: It is impressive to have tools that can make sense of this complex space. It doesn’t tell us much about steps to artificial general intelligence: AlphaStar needed 200 years of gameplay, across 14 real days and 16 Google tensor processor units, to get this good. It used a portfolio of techniques, including LSTMs (good for the long-range planning stuff) and reinforcement learning. Quite impressive, AlphaStar could play using a standard PC set-up using a single GPU; it is also, I suspect, solving reasonable industrial-like problems. Starcraft looks like many types of planning-and-control systems in the real world, such as logistics or factory operations or even general planning. It also reminds us the DeepMind are masters of PR. Demis Hassabis is a great ambassador, a relatable genius, in many ways. Other firms looking to excite the tech community (and public) more broadly should take heed.
Will.i.am argues that consumers should own their data & be able to monetise it. I’m not so sure that the starting point for data should be to think about individual’s selling their data. (My comment, together with responses from The Economist editor who commissioned the piece, is available as a subscriber-only update.)
Shreya Amin with an accessible overview of why it is hard for computers to wreck a nice beach. (Recognise speech, geddit?)
Our ability to forecast certain types of weather is improving rapidly. 72-hour hurricane forecasts have become seven times more accurate over the past four-and-a-bit decades. (Important as we get more extreme weather & as more of our electrical generation will become sensitive to wind and cloud-cover.)
🔥 Endangered planet
We are in serious shit with the fundamental damage we have done to our biosphere. So each week, I'm going to remind us of our level of the CO2 in the atmosphere. As an aide memoir, at 450 parts per million, we have a 50% chance of preventing a 2° global average rise in temperature. This is a level at which things turn really, really nasty.
This week's level: 410.40ppm (12 months ago: 406ppm; 250 years ago, est: 250ppm)
Dozens of wild horses were found dead during Australia's record-breaking heatwave this week.
Think how you can reduce your emissions: less driving, trains instead of planes when possible, introducing a planet-friendly diet… There are ways. See what I decided to do this year to go carbon neutral.
Dept of social sciences
A detailed report from the Brooking's Institute looks at which regions of the US face most employment risk from automation. The conclusion is the 'heartland', mostly Trump-supporting regions.
🇸🇬 Singapore's Minister for Communications and Information unveiled a model Artificial Intelligence Governance Framework at Davos, the first detailed guidance for the private sector, coming out of Asia. The document promotes the responsible use of AI for internal governance, determining decision-making models, operations management, and CRM. Document available here, summary here.
George Soros warns against China's influence on the open internet at Davos.
Nesta takes on mapping AI governance activities worldwide, aiming to support research and policy-making with a searchable database and visualisations.
🔎 Research on fake news on Twitter during the 2016 election. Only 0.1% of Twitter users engaged in sharing fake news, mostly older, conservative types. These supersharers were exposed to 20-times as much fake news as the median user. Another useful conclusion: “immediate points of leverage to reduce the spread of misinformation could [be] to discourage users from following the most pervasive [fake news accounts]”.
A rather thought-provoking essay on how, via Habermas, Baudrillard & McLuhan: since the era of the printing press, control over the public conversation shifts from the stories the public is interested in reading on the one hand, and the story the media is interested in telling. A challenging point:
Just as the medium is not the Message, it is not inevitable that the medium become its Message either. We must remind ourselves that individuals are not powerless and impotent. Engineers can resist. Producers can boycott. Journalists can inform, rather than entertain. The demands of the market can be rejected.
The pressure to submit is overwhelming... To refuse to go along is to be replaced. Everyone and everything is fungible.
Short morsels to appear smart at dinner parties
😮 Scientists link chronic gum disease to Alzheimer's. This is massive. I recommend regular interdental brushing!
✨ Fusion power could power our grids sooner than many would expect.
What is the future of CERN's large hadron collider? (This essay sparked a mini-backlash from many physicists.)
📺 TV is not dead yet, and it might break Google and Facebook's stronghold on advertising.
🗞️ Journalism professor, Jeremy Littau, has a great tweetstorm chronicling the decline of newspaper advertising.
Friendly fraud: Facebook’s term for allowing game developers to bill unwitting children thousands of dollars. 😳
Authorities in Chinese Hebei province created an app that notifies you when you're in proximity to someone in debt.
Here is a smartphone without ports or buttons
🔋 Looking beyond lithium as Li-ion battery nears its development limits.
Thank you! We had a cracking start to our gentle roll-out of the Member’s tier of Exponential View. It has been wonderful to welcome so many supportive members.
Included in the membership are the regular live briefings I will be hosting with a world expert (and in some cases, the world expert). Coming up:
Friday, February 1, Phil Fersht, CEO of HfS Research: How will automation in the enterprise play out and reshape companies? What will its impact on jobs be?
Friday, February 15, Rumman Chowdhury, data scientist and the Responsible AI Lead at Accenture: Ethical artificial intelligence: How should we frame the problem?
Friday, March 1, Horace Dediu, one of the world’s most renowned business analysts and experts on complex data analysis: How is micromobility transforming urban transit? What is the future of cars?
The early discount (about 27%, the equivalent of about £12 or $15 a month when paid annually) will run until February 4th so please join the rest of our early members, and sign-up now.
We also have special discounts for academics and students. Email firstname.lastname@example.org to find out more.
Have a great week!
P.S. The sharing buttons on our new platform are not as easy to find as before. So if you want to share with friends, hit this for Twitter and this for LinkedIn, or just forward the email with a personal note.
What you are up to—notes from EV readers
Stefano Zorzi's essay about the relationship between technology and nationalism. (Worth mulling over.)
Dan Howlett questions: what are the problems underpinning open-source usage in the enterprise context?
Jan Erik Solem's Mapillary just put 186 million map features (like utility poles, manholes, CCTVs, and benches) on the global map, using nothing but street-level images and computer vision.
Tim Gordon opens the Best Practices AI database of 1000+ case studies across more than 60 countries for EV readers this weekend.
Lydia Kostopoulos' optimistic outlook on the future of tech & society: why it's a wonderful time to be alive.
Ian Hogarth joins the UCL's Institute of Innovation and Public Purpose as the Visiting Professor. Congrats!
Gemma Milne interviews Azeem on what 2019 might hold for European AI startups.
Ben Dellot of the Royal Society of Arts has published A Fieldguide to the Future of Work.
Davos special: Watch reader Heather Long ask Michael Dell about whether he would accept higher taxes. (And listen as he gets pwned by economist Erik Brynjolfsson).
Just a reminder: If you’re up to something interesting, let us know. If we don’t know, we can’t share! Email email@example.com.
This week’s Exponential View has been supported by our partner: Ocean Protocol.
Ocean Protocol is kickstarting a Data Economy by breaking down data silos and equalizing access to data for all.
Learn about the Ocean Protocol software stack and how it's relevant to you: docs.oceanprotocol.com