🔮 EVs on the up; Google‘s paywall; software’s fatal flaw; creative walking, long Covid and Amazon’s ”AI” ++ #468

and

Apr 07, 2024

Hi, I’m Azeem Azhar. In this week’s edition, we explore what the trajectory of the electric vehicle market can teach us about the adoption of exponential technologies.

And the rest of today’s issue1:

Need to know: Bad bug
The recent accidental malware discovery uncovers a massive threat to digital public goods.
Today in data: Americans give ChatGPT a go
20% of Americans have now used ChatGPT at work.
Opinion: Lithium FTW
There was never going to be a shortage of lithium. And there isn’t.
AI Practice: Bye-bye, prompt engineering
Hello, conversation engineering.

Sunday chart: A bump in the road

The electric vehicle (EV) market continues to gain momentum, with global EV sales reaching 20% market share at the end of 2023. Thirty-one countries have passed the 5% adoption tipping point. While recent quarterly sales dips from industry giants Tesla and BYD might raise eyebrows, these fluctuations are better understood as the growing pains of individual companies navigating an increasingly competitive and maturing market. Both Tesla and BYD’s struggles can be partly attributed to the intense competition in the Chinese market, where there are over 129 EV brands. Last year, price competition in China gave both firms a temporary fillip, now showing up as a decline in this year’s sales.

This rapid improvement in affordability is evident in the narrowing price gap between EVs and traditional vehicles: just a year ago, the average new EV sold in the US commanded a hefty $10k premium over the market average; now, that difference has shrunk to a mere $5k. The total cost of ownership of EVs is close to competitive with or lower than that of ICE vehicles. And EVs remain on a steep experience curve with prices free to fall much further. Their lifecycle carbon emissions are lower, too. (Although, in the US, buyer satisfaction with EVs is marginally lower than satisfaction amongst those buying ICE vehicles.2)

The rapid freefall of battery pricing is helping to address range anxiety. Xiaomi’s recently unveiled SU7, a $30k Porsche clone, shatters expectations with an astonishing 500 miles of range, far surpassing the average US buyer’s demand of 300 miles. This year, car makers will offer 30 different models with a range in excess of 300 miles. (For more details on battery pricing, see my recent commentary.)

Painting the recent sales dip as a death knell for the EV industry is a gross mischaracterisation. An understandable response for incumbents who have fallen behind on this industry transformation is to point out why the market can’t change. Consider comments from the boss of Stellantis or Toyota’s Gill Pratt. But put them in the context of Steve Ballmer’s response to the iPhone.

When markets develop, there are often wobbles. It’s an error to think turbulence is a change in the flight plan.

Key reads

Millions relying on one. A Microsoft engineer has serendipitously discovered a highly sophisticated attempt to introduce malicious code into XZ Utils, a widely used open-source compression utility. Had it gone undetected, this malware would have been bundled in upcoming Linux distributions, which power the majority of the world’s servers, potentially leading to widespread security compromises. This attempt was part of a highly sophisticated, long-term campaign originating in 2021, likely by nation-state threat actors. The attackers exploited a systemic vulnerability in the open-source ecosystem by targeting the utility’s sole maintainer through social engineering. This incident highlights a critical issue: how can we protect digital public goods, often built on the shoulders of individual contributors, from deliberate contamination?

See also: This deterioration in the verification process is also evident in other areas. For instance, in the actions of large companies like Alibaba — the company has published source code incorporating a fake software package hallucinated by AI models, presenting a potentially open path for malware to enter.

Searching for profits. Tech titan Google is mulling charging for its AI-powered search, a seismic shift that would see one of its core offerings hidden behind a paywall for the first time. The impetus for this potential pivot is twofold: the hefty computational costs of LLMs compared to traditional search, and the uncertainty surrounding advertising revenue in an LLM-based search paradigm. Google may be eyeing a model similar to Perplexity, where the base search is free but users can pay for access to superior models (listen to my conversation with Perplexity co-founder Aravind Srinivas). As I pointed out last year, GPT represents a tidal wave poised to engulf the search landscape. Google’s $60 billion search ad business is powerful and growing but not immune to competition.

See also: The FT’s deep dive into how Google’s slow embrace of generative AI risks losing ground to competitors and exposes internal cultural and organisational challenges.

An immunological response. Paul Triolo’s overview highlights the resilience of China’s semiconductor industry despite US export controls. China is seeking self-sufficiency in advanced chips and manufacturing equipment and has made great strides. They can, for instance, domestically produce the Kirin 9000 chip that goes into Huawei’s latest smartphone. As I suspected in 2020, “I think it’s highly unlikely that the US can stop China developing a semiconductor industry.” As David Goldman points out, China’s semiconductor industry could potentially diverge from silicon, as Huawei invests billions in materials science research. China saw 80,000 material engineers graduate in 2022 compared to the US’s 3,400, so its potential is unmistakable.

Also, Alibaba’s Joe Tsai reckons the top Chinese firms are “two years” behind the US AI labs.

The Exponential View advantage

As a reader of Exponential View, you’re already ahead. Here’s an opportunity to share this advantage with your entire organisation. I have only two remaining keynote speaking slots open in May and June, and I’d prefer to bring my insights to you, your team or your clients. My keynotes are designed to deliver the clarity and deep expertise on AI and the forces shaping our future that leaders need more than ever — so don’t hesitate to reach out.

Inquire now

Newsreel

Microsoft unveils $100 billion plan for “Stargate” AI supercomputer.
New York City announces a permitting process for autonomous vehicles testing — albeit under the vigilant gaze of human safety drivers.
Research organisations and scientists that have received funding from the FTX Future Fund face the prospect of returning the funds to avoid possible legal action.
Anthropic researchers discover a new “jailbreaking” technique called “many-shot jailbreaking”. This can force LLMs to produce harmful responses by exploiting their increasingly large context windows.
OpenAI deems its voice-cloning tool too risky for general release.
Taiwan quake puts the world’s most advanced chips at risk.
Apple’s 3B parameter LLM, ReaLM, outperforms GPT-4 in efficiently understanding natural language references to visual context, making it practical to implement conversational assistants on devices like smartphones.

Data

In Q1, 37 startups crossed the billion-dollar valuation mark — the highest quarterly figure in a year and 48% higher than Q4 2023.

The average information transfer rate is around 39 bits per second for all human languages.

According to the Pew Research Center, 20% of Americans have now used ChatGPT at work as of February 2024, up from 8% in March 2023.

In just two years, fossil fuel electricity in the EU has seen a significant decline, with its market share dropping from 41.1% to 27.6%, a reduction of one-third.

Short morsels to appear smart at dinner parties

👁️ Amazon Fresh’s “Just Walk Out” AI system actually relies on 1,000 manual trackers in India.

🇨🇳 How China’s industrial overcapacity reveals two competing visions of the world. Excellent essay by

High Capacity

🚶‍♀️ A study suggests that walking not only boosts your mood but also significantly enhances creativity.

🧪 A new technique using DNA brings us closer to transforming harmful carbon dioxide into valuable products.

🧊 Six centuries ago, Native Americans embraced decentralisation as the planet transitioned to the Little Ice Age. Could this be a blueprint for our own adaptation to climate change?

😷 With no cure in sight, long-Covid patients are catalysing a reevaluation of research strategies.

End note

I have been reading Daniel Yergin’s The Prize, which discusses the origin and development of the oil industry. It’s meticulous, remarkable in detail, and relevant to the mechanics and machinations we see in the tech industry today. It’s not the shortest read, at roughly 800 pages of closely typeset pages but it is instructive. The parallels are hard to avoid — and recognising that we’ve seen these patterns before helps me, at least, make sense of some of today’s Magnificent Seven dynamics.

I took a short detour to read Annie Jacobsen’s Nuclear War: A Scenario. It is a hair-raising unpicking of the genesis of American nuclear policy, the logic behind MAD and, most chillingly, what might happen in the event of nuclear escalation. Like Yergin’s book, it is a page-turner. But having only a couple hundred pages, you can get through it in an evening.

Cheers,

What you’re up to — community updates

President Biden has nominated Michael Sulmeyer as the inaugural assistant secretary of defense for cyber policy.
Konrad Kollnig has been granted funding to continue his research on using technology to support AI regulation. You can find out more about his project here.
Maria Axente has co-authored “Human-Centered AI: A Multidisciplinary Perspective for Policy-Makers, Auditors, and Users” — a U7 collaborative project bringing together experts from academia and industry to advance human-centric AI. She is inviting the community to the book launch.

Share your updates with EV readers by telling us what you’re up to here.

We saw and liked a similar introduction to the newsletter in another great publication, Bankless, so we thought we’d give it a try. Thanks for the inspiration Ryan and team!

This hasn’t been my personal experience.

Exponential View