🔮 AI cloud capture; asymmetric warfare; 34 authors; code quality; big PDFs ++ #459
An insider’s guide to AI and exponential technologies
Hi, I’m Azeem Azhar. I advise governments, some of the world’s largest firms, and investors on how to make sense of our exponential future. Every Sunday, I share my view on AI and other exponential technologies in this newsletter.
Latest posts
Here’s what you missed if you’re not a member:
Sunday chart: $100bn (cloud) baby
AWS is now at a $100bn run rate. It is also highly profitable, contributing 76% of all Amazon’s operating profits over the past decade. Azure became Microsoft’s largest product line in terms of revenue in 2019, and that business still grew 30% year-on-year. Google, the smallest of the big cloud providers, clocked up a 25% annualised growth in the fourth quarter. Overall these three hyperscalers added $15bn in new annual recurring revenue in the fourth quarter of last year.
Every time a company announces its intention to invest in AI (whether generative or some other flavour), these workloads must run somewhere. For mainstream businesses, that somewhere will often be the cloud; despite rising costs, only the bravest or most capable can take their infrastructure back in-house.
There is a long-term secular trend for companies to do more in silico. Today, we might think of AI workloads as being about cost optimisations and efficiencies. But increasingly, the real value-added work of a company (whether it is route planning, pricing analytics, research and development) will run on demanding ML models. The clouds will continue to rise.
Key reads
Silicon Valley despots. Adrienne LaFrance cautions against a distinct ideology that is blossoming in Silicon Valley: techno-authoritarianism. This system allows tech companies to dictate the rules and cultural norms of the digital world, driven not by the public good, but by their bottom line. Technology has the potential to be a force for good, yet this is not an intrinsic attribute. Rebecca Solnit draws an incisive critique of Silicon Valley’s power and its dystopian impact on San Francisco:
We were the left edge of America, a refuge from some of its brutalities and conformities, a sanctuary for dissidents and misfits and a laboratory for new ideas. We’re still that lab, but we’re no longer an edge; we’re a global power centre, and what issues from here – including a new super-elite – shapes the world in increasingly disturbing ways.
Collaborative fiction. EV reader Ludwig Siegele spotlights the resurgence of collaborative writing, exemplified by the novel Fourteen Days, which had 34 author contributions. This trend is rooted in historical practices but has gained momentum with modern technologies. For example, Google Docs allows real-time editing and contributions by multiple authors, while GitHub, originally for coding, is now used for co-authoring books, facilitating a more cohesive and dynamic writing process. And of course, AI now plays a role in the creative process too. This phenomenon is exemplified by Rie Kudan (referenced in EV #457), a renowned Japanese novelist who said ChatGPT was responsible for about 5% of her award-winning novel Tokyo-to Dojo-to.
Asymmetric evolution. Asymmetry is a key factor in modern warfare — Iranian-made Shahed-136 self-exploding drones cost only $20,000 to make, while the missile to intercept them, like the US SM-2 fired near Suez, can cost up to $2.4 million. This punchy ROI has led to widespread investment in drones. For example, Middle Eastern countries, excluding Israel, invested over $1.5 billion (about $5 per person in the US) on drone warfare. However, the invasion of Ukraine has shown that military advantages soon prompt the innovation of tools and tactics to counter them. Ukrainian drones had become much less successful (around a 30% success rate for the best drone teams, dropping for the average group) because of electronic warfare. This has resulted in an accelerating rate of innovation - a complete system upgrade every three months, from changing radio links to new frame designs, different sensor packages, and new types of warheads.
See also:
LLMs in wargame scenarios tend to escalate arms race dynamics — and in rare cases deploy nuclear weapons.
Top US news outlets, predominantly left-leaning, block AI data collection to protect content, while right-wing outlets like NewsMax and Breitbart allow it. This might result in LLMs spreading views that favour conservatism.
Starlink satellites represent a paradigm shift in global communications, from traditional cell phone towers to a new era of orbital network connectivity.
Market data
Last year, 40% of China’s GDP growth came from clean energy.
People with double majors have a significant shield against earnings shocks, reducing their impact by 56%.
The projected rate of code churn, which measures the percentage of lines reverted or updated within two weeks of creation, is expected to double in 2024 relative to the pre-AI levels of 2021, according to an analysis by GitClear.
Bard Pro ranks second behind GPT-4 Turbo on Chatbot arena, an ELO-based ranking system based on human preferences for LLM outputs. Mistral has also had a model leaked that reportedly nears GPT-4 level performance. It appears GPT-4 is close to being overtaken, especially when Bard Ultra is released. Now it’s White’s move.
Short morsels to appear smart at dinner parties
🔎 The new (generated) internet? Browser company Arc has released a search app for the iPhone that creates your own custom website based on your search.
🗺️ On the complexity limits of fictional worlds.
🍏 Apple’s theatrical packaging philosophy.
🔍 Bismark Omit leafage buck bank… The 19th-century silk dress cryptogram has finally been translated, but has it been solved?
🦾 A robot can now read braille twice as fast as humans.
Partner with Exponential View
Sponsoring Exponential View is the best way to reach a global audience of executives, founders, investors and policy-makers — while supporting your favourite newsletter. Get in touch to explore how we can work together.
End note
I’m eager to hear if any readers have received their Apple Vision Pro. If you have, drop your impressions in the comments.
🍏 How Apple‘s Vision Pro can succeed
We have to think of the Vision Pro (or VP, as I’ll refer to it) less as a device and more as a computing platform. What it means for a computing platform to succeed is very different from a product’s success. A product can succeed by large numbers of sales alone, driven by whatever spirits moved the consumer: think of pet rocks and fidget spinners.
Cheers
A
What you’re up to — community updates
- published his annual presentation on the state of decarbonisation.
Alena and Erik Brynjolfsson theorise that AI may make markets less rational.
Aisling Carlson wrote about the importance of catalytic funding for climate tech solutions.
Mark Cridge launched a crowdfunding campaign for the London National Park City Visitors Centre.
Kenneth Cukier of The Economist is giving a talk on AI and God on Tuesday, February 6th at the Quaker Meeting House in Wandsworth, London.
Marko Ahtisaari is building a new House of Sound in Finland, focused on sound art and exhibitions on the physics, neuroscience, human, social, and environmental impacts of sound.
Tobias Peggs, head of Square Roots Indoor Farms, an ag-tech startup Square Roots Indoor Farms, has announced a new program to grow plants ‘in the dark’.
- has recently launched a Substack on the intersection of law, geopolitics and technology.
Share your updates with EV readers by telling us what you’re up to here.
I had the opportunity to try one for a couple of minutes. It has some qualities such as a great screen and an intuitive UI once you got used to it - pinching is not what my fingers are used to after a decade of iPhone. But right now, it’s far to heavy, no way I‘d sit on the couch and watch the latest Scorsese or do my emails. Maybe a Simpsons episode. Having to carry a cable bound separate battery also doesn’t feel apple-ish. So for me, right now, it’s a pretty expensive Version 1 gadget which I probably won’t buy. I guess it will come down finding the killer use case (reading emails or watching Dinos are not) and whether Apple finds ways to optimize weight.
What does "AWS is now at a $100bn run rate" mean? The graph shows $25bn.