🔮 Self-driving distraction; mature AI; supply chains, eyes & education++ #393
Self-driving could be a pointless distraction for improving the environmental and human impact of transport
Hi, I’m Azeem Azhar. I convene Exponential View to help us understand how our societies and political economy will change under the force of rapidly accelerating technologies. The Sunday edition of Exponential View is a thoughtful curation of evolving thinking across business, technology, economy and culture. Sign up for the weekly insights here.
In today’s edition:
The distraction of self-driving vehicles,
Machine vision & medicine,
Breaking tech supply chains.
The near future
🛣️ End of the road
In 2016, the then-CEO of Lyft shared this chart. 2022 would be a year dominated by autonomous driving.

Now Max Chalkin analyses the disappointing trajectory of full-self driving efforts: $100bn invested and little to show. Self-driving pioneer, Anthony Levandowski who cofounded Waymo, has retreated to building autonomous trucks constrained to industrial sites. He reckons that is the most complex use-case the technology can deliver in the near future. (Amazon has also abandoned its autonomous delivery robot programme.)
Why it matters: Self-driving could be a pointless distraction for improving the environmental and human impact of transport. It takes attention away from micromobility, better urban infrastructure and other strategies to improve the safety, pollution, climate, equity and economic returns of this sector.
🤖The AIs have it
Machine vision systems are able to look at retinal scans to predict heart attack and stroke risk, as well as the early onset of Alzheimer’s. Such patterns were invisible to human researchers. As Eric Topol writes, “the power of machine eyes extends to virtually all forms of medical images” into the domains of oncology, gastroenterology and more.
Another machine vision system can estimate the energy efficiency of buildings using images from Google Street View. [PDF of the paper]
Why it matters: Machine vision is one of the more mature applications of deep learning, making it readily accessible to researchers and industry. Data sets are increasingly easy to come by, and the computational cost is far from the state-of-the-art experiments in super-large AI models. This bodes for even more rapid and widespread uptake.
🗺️ Running in the shadows1
The Biden administration announced severe new rules on exporting advanced technologies (and the tooling that might build them) to China. The goal is to hobble the Chinese state’s access to the most sophisticated chip tech. The rules explicitly reference the civil-military fusion approach Beijing takes and the leakage of such technologies into the military sphere. Dylan Patel’s analysis is exceptional.
Indian factory lines for Apple’s latest iPhone, the 14, have already started to turn in the sub-continent only a few weeks after Chinese production lines spun up. This is a boost for Apple, as it looks to diversify manufacturing beyond China, and for India it is a hope to benefit from “China+N” sourcing.
However, Bloomberg reckons that the process of multi-shoring will be slow: 8 years to move a tenth of Apple’s production from China. The decades (and tens of billions of dollars) Western firms have invested in capacity in China, as well as the depth of the supply chain and expertise that has emerged there, are making it a hard break up. I’m not so sure it’ll be as slow as that.
Why it matters: The unwinding of the previous model of globalisation is going to happen, despite huge investments. With national security (rather than merely commercial resilience or economic competition) becoming a primary driver, the speed could accelerate.
Weekly Commentary: The machine that builds itself
In this week’s commentary, I look at recent progress in AI research. Might we be seeing recursive self-improvement in AI systems?
Dept of our climate future
In every Sunday edition, we track key metrics that tell us a little about our shared climate future. Our member, Marshall Kirkpatrick, takes the time to curate a view of our current climate status in this segment every week, and you can read Marshall’s curation below.
The one thing about our climate future you should know this week: New analyses from multiple quarters now argue that the transition to a green economy is crossing a fundamental threshold; one which “definitively changes the narrative from risk mitigation to opportunity capture.” That’s the big take-away for large corporations from the US Inflation Reduction Act, according to a maddeningly-unavailable report from Credit Suisse quoted by Robinson Meyer in The Atlantic this week. According to Meyer, the report focuses on the Act’s uncapped tax benefits and the follow-on spending in the private sector that “could send total climate spending across the [US] economy to roughly $1.7 trillion over the next 10 years.” That’s multiple times larger than the official estimated impact of the IRA. This analysis is complimentary with Azeem’s recent essay The Negative Cost of Net Zero, which points to an exponential growth of solar, wind, batteries and electrolysers, tied to dramatic cost declines, made possible by a virtuous cycle of continuous expansion of capacity and the concomitant growth-driver of learning. The question of narrative should concern all but the most triumphant materialists; this next “gold rush” must not play out like the last one. If we’re to get through what the Cascade Institute recently called a global polycrisis, it will likely require an effective mega-portfolio of interconnected solutions.
In other news about our climate future:
Chemical engineers at the University of Melbourne say they’ve developed a direct air electrolyser that can skip the water and extract green hydrogen from the air, at humidity as low as 4%.
The Chinese Academy of Sciences has set a new world record with a 100 MW compressed air energy storage plant, beating the previous record held by a 64 MW plant 1000 meters underground in a decommissioned salt mine. Compressed air is a lithium-free way to store energy that China says will account for nearly a quarter of its energy storage needs by 2030.
The venerable Rocky Mountain Institute has released two large, newly available reports on the most effective and equitable ways to transition local economies away from coal and to green energy.
The state of Vermont has followed California and launched a trade-in program, where older ICE vehicles can be exchanged for a voucher covering part of the cost of a new EV. “Every single US state should be running an incentive program like this, Electrek writes.
Market data
Tiktok’s operating losses triple to $7bn. Its net loss for 2021 hit $84.9bn!
Meta has chucked $70bn at the Metaverse.
Decentraland, the Web3 metaverse play, has 38 daily active users. Its token, MANA, has a total capitalisation of $1.3bn.
Overall revenue generated by the Web3 infrastructure sector declined 67% QoQ to $5.6 million in the third quarter.
Rwanda estimates it will cost $11bn through 2030 to attract new investments to fight the environmental impacts of climate change.
Short morsels to appear smart while resting your eyes
🔋Tony Fadell, co-creator of the iPhone, isn’t worried about the EU mandate that will see the iPhone junking Lightning connectors in favour of USB-C. I spoke with Tony on the Exponential View Podcast earlier this year, and you can listen to our conversation here.
📉Instagram is about to be ruined by more ads. Also, don’t travel where social media tells you to.
🧫 Dormant spores decide to wake up by counting in their sleep using electricity.
🩸Scientists have discovered an ultra-rare, new set of blood groups.
👓A surprising reason for the growing eyesight crisis: education.
🏘️ OpenDoor’s house price predictions have gone awry, as changes in fundamentals play havoc with its models.
⚔️Timothy Snyder, a scholar of Eastern Europe, provides a stunning analysis of how and why the Russian invasion of Ukraine will end. (And also, how the 25-year-old Himars weapon system is playing its part in the war.)
End note
We are testing some new ways of presenting Sunday’s newsletter and the weekly commentary. Expect a few iterations in coming weeks. Comments are open for your feedback!
📅 Member event: EV member, Claudia Chwalisz, recently launched DemocracyNext, an international non-profit, non-partisan research and action institute dedicated to building institutions for the next democratic paradigm. She will present her new venture and discuss with EV members how participation can help us reclaim democracy on October 20, 5pm BST | 12pm ET. The event is open to paying members of Exponential View, and members can RSVP here.
What you’re up to – notes from EV readers
Pascal Finette published a new Disrupt Disruption podcast episode discussing innovations for complex, interdependent environments with Natasha Gedge.
Charles Yang wrote about his time at ARPA-E working on quantum computing, and real-world examples of path criticality for technology development.
Peter Schwarz published an op-ed in the Chicago Tribune about the politics of capital letters.
Robert Hackers laid out a strategy for academic research in the 21st century: guidance for early career researchers.
Gianni Giacomelli published his monthly digest of stories about collective intelligence, how tech augments it, and their impact on the future of work.
Matti Bispham is giving a talk entitled “CatchM3ifuKan - Detecting Command and Control Techniques Up and Down the Networking Stack with Streaming Statistical and Machine Learning Techniques” at Zeek Week on October 13th.
To share your projects and updates, fill out your details here. Because of space constraints, we prioritise updates from paying members and startups I have invested in. (You can become the former by subscribing, if you have not already, and the latter by getting an intro to me via a trusted contact.)
With apologies to Fleetwood Mac.
The (dubious tech) HOW of completely independent self-driving cars is likely a distraction, but the WHY of it could still generate benefits. For instance, user centered design of self-driving vehicles can lead to seamless public transport where a digital assistant plans every minute of your commute. And/or 5G enables remote control of a vehicle that is partially self-driven. The end result for the user still feels autonomous, but the operating mechanism is more tractable.
The Bloomberg piece on driverless cars misses the point on how progress happens - i.e. in fits and starts, with many bumps. The overinflated valuations aren't a feature of this tech, but of almost any tech over the past 5 years. And $100bn sounds like a lot till you consider that the US spends $4tn annually on healthcare and wastes a big chunk of it. Or that global automotive market is circa $3tn so this is about 3% of one year's revenue that has so far invested in disruptive tech. And finally, isn't this why exponential markets confuse people? For a long time it feels like nothing's happening, but then when things happen, they seem to happen very quickly. Yes, the optimism was misplaced. But 'going nowhere' seems unduly pessimistic, and centred around Levandowski's rise and fall.