🔮 Apple’s billion users; against knowledge; unsustainable AI; the autonomous economy; wealth, stamina and pregnancy++ #221
An intellectual feast in London next week, join me
Exponential View
Azeem Azhar’s Weekly Wondermissive: Future, Tech & Society
This issue has been supported by our partner: Ocean Protocol.
See how Ocean Protocol is helping to bring innovation to patient care.
You are reading the free edition of Exponential View
Please do me a favour and help with a couple of things:
First, to prevent future mass-spam problems, star this in your inbox and add azeem.azhar@exponentialview.co to your contacts—especially if you’re using your corporate account;
Second, we’re starting to prepare for the fourth season of the Exponential View podcast; if you haven’t listened to my conversations thus far, you really should. What would particularly help is if you could 1) subscribe, 2) give us a 5-star rating and, 3) leave a nice review on your app of choice: Apple podcasts, Spotify, Overcast, Google or Breaker.
This won’t take more than a minute or two, but means a lot to me & my team. Thanks!
Dept of the near future
🍏 Sometimes it’s worth reminding ourselves just how big Apple has become. Roughly a billion people around the world are using over 1.4 billion Apple devices. Neil Cybart argues that, with opportunities for upselling and new services revenues, ‘the company has reached a level of ecosystem strength that still hasn’t been fully digested by the marketplace.’
💯 Exploring the autonomous economy. My conversation with complexity economist and theorist of technology, W. Brian Arthur. Recommended.
🇺🇸🇨🇳 The looming 100-year US-China conflict. Martin Wolf writes in the FT about the US-China rivalry, and that countering China is becoming an ‘organising principle of US economic, foreign and security policies.’ Wolf argues that it doesn’t need to be this way, and that China’s rise can be managed through a blend of cooperation and competition. Grim reading.
The tragedy in what is now happening is that the administration is simultaneously launching a conflict between the two powers, attacking its allies and destroying the institutions of the postwar US-led order. Today’s attack on China is the wrong war, fought in the wrong way, on the wrong terrain. Alas, this is where we now are.
💣 The arrival of Amazon’s HQ2 in Crystal City, the ‘cradle of US government contracting’, marks a significant leap forward for the burgeoning military-AI complex. Amazon’s decision to metaphorically and literally move closer to the Pentagon comes in the midst of a growing tech Cold War between the US and China. The US continues to campaign aggressively against Huawei around the world and the Chinese military is reportedly considering replacing Windows with a home-grown OS. (See EV#219 for more analysis of the US-China spat.)
🕵️ The balance of power in global espionage is shifting, writes Edward Lucas in Foreign Policy. Closed societies like China, Iran and Russia now have the advantage over more open ones, and a gulf is opening up between a handful of countries with the most advanced technological capabilities and the rest of the world; furthermore, the boundaries between public and private sector intelligence work are being blurred.
🤔 Knowledge is a Stone Age concept we should leave behind, argues David Papineau. Quite a challenging and interesting piece with growing relevance as we increasingly use machine learning systems which are stochastic rather than deterministic in their outcomes. ‘We keep favouring weaker direct evidence over good statistics [...] Time and again, we’re more ready to act on information that fits the archaic stereotype of directly caused knowledge than on good statistics. But it’s all a bad idea.’
Climate catastrophe: 414.19ppm | 4,000 days
Each week, we’re going to remind you of the CO2 levels in the atmosphere.
The latest measurement (as of June 6): 414.19 ppm; 12 months ago: 409ppm; 50 years ago: 326.66ppm; 250 years ago, est: 250ppm. Share this reminder with your community by forwarding this email or tweeting this.
We’ve passed the seasonal peak for readings of CO2 at Mauna Loa. As our plants green, particularly in Siberia, they will suck some of the CO2 out of the atmosphere. Focus on the delta over the past 12 months. The delta itself continues to increase.
I’m going to introduce another metric, helpfully sparked by an email from EV reader Gavin Starks: on current trends, we are roughly 4,000 days from the 450ppm threshold. I think it is another number to help us focus the mind. Starting today, June 10th, I’ll count down from the 4,000 days every week in Exponential View. I hope it will help focus our collective wills. 4,000 days is an approximation. But, I hope you find the countdown clock useful. Action, in this case, is better than inaction.
🇫🇮🌿 Finland’s new government has pledged the nation will become carbon neutral by 2035.
Phillip Hammond, Britain’s chancellor, has said that it will cost £1 trillion to get Britain to carbon zero by 2050. The suggestion was that this would crowd out other investment priorities (although it’s hard to see which could be more important). The thinking is muddle-headed. Many of the technologies that are required for decarbonisation are enjoying periods of exponential improvement (renewables, storage costs, AI systems to manage grids). At current rates, over the next 30 years, the UK will spend £300bn or so subsidising fossil fuels. The UK bank rescue of 2008 was a package close to £500bn. Our economies can afford it.
More than 21m homes in the UK alone depend on natural gas for heating and/or cooking. Replacing that infrastructure with zero-carbon solutions would result in a remarkable large-scale effort. For the sake of argument, say it takes 8 person-days to replace the gas heating infrastructure in a home with something that is zero carbon. That is about triple the time it takes to install a new gas boiler and would result in 150m days of work to physically do the replacement or about 1 million years of work. This alone could create 200,000 highly skilled installation jobs for five years, let alone the ancillary roles around it.
Driving investment, incentives, taxation and research into decarbonisation would surely bring the cost of the tech technology down while driving up its capabilities. Concomitantly, timescales would compress as we propel to a higher energy level of a renewable, sustainable economy. There might be some short-term pain in this process of decarbonisation which can be borne progressively by the biggest historical and current emitters and those best able to afford it. The problem here is one of imagination. To paraphrase Gramsci, we should have a ‘pessimism of the intellect’ but demand the ‘optimism of the will.’
Dept of artificial intelligence
Training neural networks are exceptionally energy inefficient. Take Transformer, a neural net for natural language understanding. Training it to a state-of-the-art level would require 656MWh of energy. Assuming you did this on Amazon Web Services, this would have a carbon footprint of around 283,000kg of CO2E (equivalent to five times the lifetime emissions of an average car).
So many fascinating ideas:
The human brain, which is competitive with one of these Nets at this sort of task, require about 20W. Assuming you are awake 16 hours a day, and live for 80 years, your brain will require about 9MWh.
Data centres need to move to renewables rapidly—and we need a carbon tax.
There are nuances here. Inferencing (that is getting the system to make the correct predictions, rather than training it to make them) is much less energy-intensive. And well-trained algorithms may well reduce carbon loads through offsetting other behaviours. For example, imagine a version of the Transformer net that can translate between spoken-language-pairs in real time. That might help eliminate tens of thousands of flights every year--as teams can more easily collaborate across regions and languages.
The UK’s Information Commissioner’s Office and the Alan Turing Institute have been researching what explaining AI decisions means in practice. They found that context is key. What people seek from an explanation varies depending on the setting. Consumers want explanations in the case of recruitment or legal algorithms but, in healthcare, they care to know that a decision was accurate even if it wasn’t explicable.
⚖️ Scribes attack printing press. Not quite, but a new French law prevents firms from ‘publicly revealing the pattern of judges’ behaviour in relation to court decisions.’ Such decision-making is relatively easy to benchmark and model—and could lead to more even-handed judicial decision-making. (See also the essay on knowledge above.)
The AI data supply chain is an opaque one. Firms often don’t tell users how their data will be used—and often lose sight of where it ultimately is used. This is going to be an interesting battleground for legislators and regulators in the coming years. Shoshana Zuboff and I will touch on this issue in an upcoming episode of my podcast. (This issue is also impacting ‘data-for-good’ projects, reports Amy Maxmen in Nature: ‘concerns are rising over the lack of consent involved; the potential for breaches of privacy, even from anonymized data sets; and the possibility of misuse by commercial or government entities interested in surveillance.’)
Megvii, the Chinese AI company behind the Face++ facial recognition technology, may be the next company to find itself caught in the crosshairs of the China-US tech war. Interesting profile of the firm and its founder, Yin Qi.
Dept of the future of work
This section is only available to premium members of Exponential View.
Short morsels to appear smart at dinner parties
💰 Wealthy individuals and rich corporations are struggling to invest their piles of cash. (Excellent short read.)
Scientists say they’ve found the limits of human endurance.
🌌 NASA is throwing helicopters, and themselves, at the ground just to see how hard it would be to get good pictures of Venus. Researchers think Venus could be the key to finding other habitable worlds.
⚡ The Indian government could force ride-hailing services to go electric.
Connected fitness startup Peloton has filed to go public.
🔩 Apple is putting the screws on third-party developers in the name of user privacy.
Your gut bacteria effects how you metabolise drugs, which has important implications for medical treatment.
Scott Kupor of a16z has written a new book about venture capital and how to get it. His colleague Andrew Chen sums up some of the key topics.
🏳️ Stanford Professor Emeritus of History Barton Bernstein digs into the archival research on the bombings of Hiroshima and Nagasaki in 1945, arguing that American leaders could probably have ended the war without resorting to using the A bomb—but that, at the time, ‘there were few moral restraints left in what had become virtually a total war.’
😨 A new novel about the commerce and politics of surrogacy sounds unnervingly plausible. In ‘The Farm’, poor immigrant women carry the babies of wealthy parents.
End note
If you are in London, you should make your way to CogX, the Festival of AI, launched by (EV readers) Charlie Muirhead and Tabitha Goldstaub. Dozens of readers are speaking or otherwise involved. I am running a stage, Cutting Edge, with Libby Kinsey. We are looking at what comes next as we careen into the near future.
I tried to encapsulate some of this event in this twitter message.
As a reader, you get 25% off if you register here. Swing round to the stage and wave ‘hello’!
Cheers,
Azeem
P.S. Scroll down for some amazing news and projects by our readers!
This issue has been supported by our partner: Ocean Protocol.
Searching for the right dataset can be like looking for a needle in a haystack. With the right metadata, this search is optimized. Learn how metadata works on Ocean Protocol.
What you are up to — notes from EV readers
Congratulations to Geraint Rees for being named the Pro-Vice-Provost for artificial intelligence at the UCL.
Marija Gavrilov explores how automation of care will make us re-evaluate human empathy.
Stephan Samson’s AkademyAI offers a full scholarship to women to attend a summer AI engineering bootcamp in Barcelona. Don’t miss this opportunity!
Gabriele Mazzini maps out the the intersections between AI and the EU law.
Jaimie Boyd, director of Canada’s Open Government, hosted the 6th Global Summit of the Open Government Partnership in Ottawa. All sessions have been recorded and are available here.
Arif Khan’s discusses SingularityNET’s new partnership with Domino Pizza and how they’re using AI to enable efficient supply chain operations.
Julia Ross shares an exciting company builder program for entrepreneurs who want to create and implement solutions for the elderly. Apply by the end of June here.
Eric Beinhocker discusses the Invisible Hand and whether greed is good on BBC Radio 4.
To share your project and news with the EV community, email marija@exponentialview.co