🔮 China’s car charge; the solar cycle; dominant Google; neutrinos and extinct humans ++ #438
Your insider guide to AI and exponential technologies
Hi, I’m Azeem Azhar. As a global expert on exponential technologies, 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 developments that I think you should know about in this newsletter.
Latest posts
If you’re not a subscriber, here’s what you missed recently:
🎨 Thanks to our sponsor Masterworks, an investing platform that enables everyday people to invest in multimillion dollar paintings by artists like Banksy and Basquiat.
Sunday chart: The Chinese (Car) Century
China has potentially already reached peak gasoline demand, two years ahead of schedule, according to the country’s top oil refiner Sinopec. BloombergNEF predicted it to peak next year. So, what happened? Robust electric vehicle sales. Passing this milestone reflects the vigour of China’s EV ecosystem, which has helped propel the nation to its current status as the world’s top car exporter.
China’s ascent is no accident; the government has given unwavering regulatory and financial support for EVs for more than a decade. The latest demonstration of this commitment, earlier this year, is a tax incentive package for EVs worth $72.3 billion, its largest ever.
Explanations of China’s success in EVs often focus on its manufacturing prowess, but the nation’s strides in autonomous driving should not be overlooked. Numerous companies have rapidly emerged, fiercely competing to roll out their autonomous-driving pilot programmes in cities across the country. The valuable data gathered from these large-scale deployments will further boost their autonomous vehicle development, making them even more competitive. Along with their cost advantage, this positions Chinese firms strongly against international automotive giants.
Despite Tesla’s current lead in autonomous driving technology, Western nations’ automotive market share is likely to decline, unless other automotive heavyweights (such as those in Europe) can catch up to the pace. But it seems unlikely that either Detroit or Deutschland will cut the mustard anytime soon. Although Biden’s administration has just ponied up $12bn to help US automakers retool their plants for EVs.
China’s dual expertise in both electrification and autonomy presages a global upheaval in the automotive landscape. Cars have been shaping economies and cultures for the past hundred years. As we embark on this “second century of the car,” many signs suggest that it will be China’s hand on the steering wheel.
🎨 Today’s edition is supported by Masterworks.
Billionaires wanted it, but 54,578 everyday investors got it first… and profited
When incredibly valuable assets come up for sale, it’s typically the wealthiest that end up taking home an amazing investment. But not always…
One platform is taking on the billionaires at their own game, fractionalising prized blue-chip artworks for its investors. In just the last few years, its investors have realised annualised net returns of 17.8%, 21.5%, 35% and more.
It’s called Masterworks. When Masterworks sells a painting — like the 16 it’s already sold — investors reap their portion of the profits.
Offerings can sell out in minutes, but Exponential View readers can skip the waitlist to join.
See important disclosures at masterworks.com/cd
Key reads
Billion dollar bAIby. OpenAI’s revenue is projected to reach a billion dollars over the next 12 months, five times the $200 million revenue it predicted for this year. Much of this surge comes from the growing corporate enthusiasm around LLMs. Notably, 40% of S&P 500 companies have mentioned AI in their latest earnings calls.
Most likely driven by these corporate customer demands, OpenAI has introduced ChatGPT Enterprise. This offering addresses a key concern for businesses: the risk of uploading sensitive data to ChatGPT. With the Enterprise edition, OpenAI assures users that its models won’t learn from client-specific data. Moreover, future updates will offer model customisation tailored to individual company data. One thing is for sure, OpenAI won’t be short of eager customers.
See also: Google has introduced its Duet AI assistant to all Workspace apps, offering features that aid in tasks like converting Google Docs outlines into slides or generating email responses.
Chip Wars. As we pointed out in our NVIDIA Chartpack, computer supply is a major bottleneck in AI. Semianalysis paints a vivid picture of two camps: the GPU-rich titans like Google, OpenAI and Meta, and the GPU-poor, which includes startups - and pretty much Europe in general. Among the GPU-rich pack, Google leads. It will likely have more TPUv5’s (a Google AI chip) than all of its major competitors’ GPUs combined. This capacity will allow Google’s engineers to iterate fast while training their new Gemini model, probably reaching 20x the pre-training compute that GPT-4 will have used by the end of next year. On top of this, Google’s new TPUv5e chip provides a direct challenge to NVIDIA, being more cost-effective than A100 and H100 GPUs for models with fewer than 200 billion parameters. Google has laid a strong foundation; let’s see what they build on top of it.
Solar cycle. The future of solar is looking sunnier than ever. According to the latest IRENA data, solar PV has had a 33% learning rate - exceptionally high, outperforming most technologies. This means that with every doubling of cumulative capacity, unit costs fall by 33%.1
Solar capacity is growing exponentially, with the doubling time accelerating in 2023 compared to previous years. It took four years for capacity to double in 2015; three years by 2018; and less than two years by 2023. (This chart shows the net new additions of solar capacity.)
This rapid growth means solar costs will likely continue decreasing due to ongoing learning effects. Plummeting costs will increase demand even further, creating a virtuous cycle. Yet while the past has been exponential, future forecasts are still resoundingly linear. These forecasts were wrong before, and they’ll likely be wrong again. What was Einstein’s definition of insanity again?
Weekly commentary
In the weekly commentary available to paying members on Monday, I’ll explore the question of how we approach global AI governance in the face of an increasingly fragmenting geopolitical and economic environment. I’ve been musing about this ahead of the AI Safety Summit being held in London in November.
Market data
Llama can program. A fine-tuned model of CodeLlama-34B achieved a score of 73.8% on the HumanEval pass@1, outperforming the GPT4 score of 67%.
Pollutants that exceed WHO guidelines decrease global life expectancy by 2.3 years. For comparison, tobacco usage reduces life expectancy by 2.2 years, and malnutrition by 1.6 years.
In the second quarter of 2023, 23 countries globally reported that more than 5% of their new vehicle sales were electric vehicles.
The first half of 2023 saw a collapse in EU fossil generation by 5% YoY, leading to the lowest output on record.
Short morsels to appear smart at dinner parties
⚛️ The first observation of neutrinos inside a particle collider.
🗞️ The power of power laws in media.
📦 Aussie cardboard drones credited with Ukraine’s destruction of Russian jet fighter. See also, drone racing is experiencing its Kasparov moment.
🧬An introduction to collective intelligence, from a biological perspective. via EV member Gianni Giacomelli
💰Are people generous when the financial stakes are high?
👶 What can adults learn from kids’ friendships?
🦣 Humankind was nearly extinct.
End note
There are definite back-to-school vibes right now and I’ve been getting my office ready for a few months of solid work. I’m really struggling to find a decent pair of in-ear headphones that have an exceptional outbound voice quality. My current cans, BeatsFit Pro, are fine for sport and listening to music but the voice/microphone quality, as
can attest, is horrendous.I’m looking for recommendations. The spec is simple: exceptional outbound voice quality on POTS calls, WhatsApp calls and Zoom, as well as decent audio quality and noise cancellation/isolation. They should connect to my iPhone via Bluetooth. I won’t use these for sports, so moderate IP ratings are fine. Ideally, in-ear. The new Jabras and Sony buds are high on my shortlist, but I’d love to hear your experiences or recommendations. Best done in the Exponential Do community which annual members can access by applying here.
Have a great week!
A
Sponsor us
To reach 260,000+ readers across our audience of executives, investors and founders, get in touch to explore sponsorship opportunities.
What you’re up to — community updates
Nathan Benaich comments on why China has no place at the UK’s AI Safety Summit.
Tomás Hernando Kofman has written a summary walkthrough of Outlive by Peter Attia which takes one hour to read. I highly recommend you read the book.
Louis Rosenberg wrote a piece for VentureBeat on the emerging dangers of ‘generative inbreeding’ to human culture.
Share your updates with EV readers by telling us what you’re up to here.
Read my book for more details on this process.
A couple questions...
1. are their projections for a reduction in the absolute number of cars and/or a reduction in the size of cars as AVs become available?
2. could something like transfer learning be part of progress on climate and other common problems? For example, perhaps Brazil or the US also want to be able to produce cheap solar panels. Do we need to go through the same learning curve for solar panels as China did? More broadly, how specific is a learning curve? Is China applng lessons from solar panel production and HSRs to manufacturing in general? If so, is this further motivation for the US to make sure we master at least some areas of manufacturing?