474: Apple Cloud Gaming, Elon Musk vs Tesla & OpenAI, LVMH, Dialup Era, Batteries, Chip Smuggling, Neuralink + Humans, and Malaria Vaccine

"I’m glad I witnessed the transition"

The price of being a sheep is boredom. The price of being a wolf is loneliness. Choose one or the other with great care.

—Hugh Macleod

🐌 💾💻🤖 Young people who were born into the digital age and have never known anything else are experiencing a ‘Back to the Future’ moment with generative AI and LLMs.

I don’t mean the capabilities, which are quite novel, but the speed!

Computers and the internet used to be really slow. I don’t mean just looking back, compared to today. We knew that dialup was slow *as we were using it*…

If all you’ve known is broadband and computers with GHz-range multi-core CPUs, SSDs, and gigabytes of RAM, watching words from a LLM appear one by one on a screen or waiting a minute for a Midjourney image to appear may give you a small idea of how *everything* used to be in the 1990s and early 2000s.

The first computer I used was a 386 DX/25 MHz with 4 megabytes of RAM and a spinning hard drive that held around 200 megabytes (yet those 25MHz were enough to run Doom, which goes to show how insanely optimized John Carmack’s assembly 3d texture-mapping code was 🏎️).

For a long time, all I could do on that computer was offline. Because I was always running out of storage, I often had to delete one game or app to install another one (using 10-20 floppy disks). Then when I was tired of it, I’d delete it and re-install something else. This process could last an hour or longer each time.

The first time I went online, it wasn’t on the internet. It was on a local bulletin board system (BBS), aka someone’s computer that you could connect to over a wired phone line using a modem. You often had to re-dial for a long time, because the line was always busy — yes, they could only handle one user at a time. Some bigger BBSes had multiple phone lines and could have a handful of users in parallel.

Everything on a BBS was monochrome text or ASCII art. The characters would appear on screen similarly to how LLMs generate tokens. The first modem I used had a speed of 2400 bauds, which rounds up to about 0.3 kilobytes per second. A few years later, my 56k modem seemed massively faster, doing 6 kilobytes per second on a good day.

You know what was the worst?

When someone else in the house wanted to make a phone call and picked up the phone, it disconnected you from that BBS or multiplayer Doom game you were playing with a friend…

I’m not trying to sound like Grandpa Simpson.

I think it’s important to remember how things were to appreciate how much progress we’ve made — and to remember that future progress is possible.

I’m glad I witnessed the transition from offline to online. I think it’s a useful experience to have had. Maybe my kids will remember the transition from pre-AI to post-AI 🤔

🚨🎙️🎬🇲🇽🦅🗣️🗣️🗣️ In case you missed it, earlier this week I released a podcast with Ed William (🇬🇧) and Dylan O’Sullivan (🇮🇪) about a film we had a lot to say about:

• 🎧 Sicario by Denis Villeneuve: Going Deep with Ed William and Dylan O'Sullivan 🎥 🎬 🎞️ 🍿

Speaking of Villeneuve, I knew he had mentioned he wanted to do ‘Dune: Messiah’ to (mostly) conclude Paul Atreides’ arc, but I wasn’t sure if that’s what he would do after ‘Dune: Part 2’.

I found this recent video about the rumored FOUR scripts he’s working on:

• Dune: Messiah

• Cleopatra

• Rendezvous With Rama (Arthur C. Clarke, classic sci-fi!)

• Sicario 3 (with the original cast back, including Emily Blunt, and Roger Deakins behind the camera!)

The video claims that Sicario 3 has a shooting start date of February 8, 2024. Next week!

Which of these would I most like to see? That’s tough. They all sound good to me.

But Rama is *very* intriguing. I have pleasant memories of that book. I remember it being very atmospheric and full of large-scale wonders, which would play to Villeneuve’s strengths.

💚 🥃 🚢⚓️ If you feel like you’re getting value from this newsletter, it means the world to me that you become a supporter to help me to keep writing it.

If you think that you’re not making a difference, that’s incorrect. Only 2% of readers are supporters (so far — you can help change that), so each one of you makes a big difference.

You’ll get access to paid editions and the private Discord, and if you only get one good idea here per year, it’ll more than pay for itself!

---

🏦 💰 Liberty Capital 💳 💴

🥾 Layoffs, layoffs, layoffs ✂️

Lately, I've been wondering if there have been more layoffs than usual or if it was just that I was noticing them more.

Or maybe it’s the *kind* of layoffs that has been different?

Anecdote vs data: The graph above surprised me. While the huge spike in 2020 makes sense, I’m surprised that what seemed (to me) like an unusually high number of layoffs is still way below the long-term trend.

There were more layoffs in 2023 than in 2022 or 2021, but it is still the year with the third-fewest layoffs since we started counting.

It makes sense when I stop and think about it. The US economy is doing well and unemployment is low.

It must be that after so many years of the tech industry hiring more people and increasing their comp, it felt very jarring to see so many layoffs. But that’s still just one corner of the larger economy.

Maybe the chart would look different if you could graph “dollars of salary lost to layoffs per year” 🤔

Via Matt Darling, h/t my friend MBI (🇧🇩🇺🇸)

Musk's AI Startup Wants to Raise $6B to Take on OpenAI 💰💰💰💰💰💰🦾🤖

There’s never a dull moment in Musk World.

On top of threatening — blackmailing? — Tesla that it will stop developing his AI/robotics ideas there if he’s not given additional compensation of around 7% shares in the company to reach 25% ownership*, Musk is apparently trying to raise big bucks for xAI:

Three people with knowledge of the talks said Musk hoped to raise as much as $6bn in fresh equity capital for xAI at a proposed valuation of $20bn. However, the people cautioned that negotiations were ongoing and that the Tesla chief was still testing investor appetite for such large sums.

For those who are worried that Musk’s interests in China make him vulnerable to leverage by potential US adversaries, this will not assuage those fears:

According to four people, these talks have included family offices in Hong Kong, the territory that is increasingly controlled by Beijing. [...]

One person said he had also targeted sovereign wealth funds in the Middle East, while others said investors in Japan and South Korea had been approached.

The biggest beneficiary of this fundraising is likely to be… Jensen Huang 😎

* To come back to this: on one hand, if the board and shareholders agree, and the extra comp is conditional on some long-term goals and KPIs that make sense and can’t easily be gamed… it’s their company and their money and may be a good deal. If Musk is that valuable to Tesla, he has leverage to extract huge comp (as he has done in the past, though that’s running into legal roadblocks) and it can still be win-win.

On the other hand, it *feels* kind of wrong. Can you imagine if Buffett had said this once in a while? “I’m the biggest shareholder and worth tens or hundreds of billions, but I want Berkshire’s board and shareholders to grant me x% of the company in comp or I may start to do some of my investing outside of Berkshire, in my PA or a separate vehicle?” What about if Bezos had done this periodically at Amazon? 🤔

Update: After writing the above, I found a quote by Musk where he says his preference is for a “dual-class stock” and that he’s “not looking for additional economics; I just want to be an effective steward of very powerful technology."

Super-voting shares probably make more sense, but it still feels a bit like blackmail to ask the company to give him extra control (which *has* big economic value, so it’s not like there would be zero economic impact) *or* he’ll stop implementing his best ideas at the company.

🔋🔋🔋 Batteries in Two Graphs 📈📈

While I’m not a big fan of the Rocky Mountain Institute (RMI) because of the damage their misguided advice has done to global energy grids, I like these graphs.

The first shows global battery sales by sector in GWh/year:

The second shows battery cost and energy density since 1990.

While these are extremely impressive human achievements, let’s be careful not to extrapolate things too far.

The multi-decade success of Moore’s Law has kind of trained us to expect exponential improvement, but it’s not a law of nature. It requires a lot of investment and ingenuity.

Computer chips are physical goods, but because the high-value silicon die is very small and very thin, the cost of materials isn’t a big factor. It’s all about R&D and upfront capital investment (think ASML machines and fancy multi-billion fabs).

While battery production can be highly automated, chemistries improved, etc, they have higher variable costs because more of their cost is going to raw materials, not just to developing IP and capital goods.

While the numbers on the graphs above are big and are growing fast, what needs to be done is gigantic. There can be bottlenecks or bumps in the road causing these curves to bend down… It’s easy to underestimate the scale of the physical infrastructure that needs to be changed to do things like electrify transportation or clean up the power grid.

🍎 🔑🔓 🎮👾Apple Un-Bans Game Streaming in App Store

Good news for gamers looking for more ways to go *pew pew pew, lasers!* and good news for Nvidia’s GeForce Now and Microsoft’s Xbox Cloud Gaming (I’m not sure I should even mention Amazon’s Luna… And Google’s Stadia is 🪦).

Apple changed its app store policy:

Developers can now submit a single app with the capability to stream all of the games offered in their catalog.

Apps will also be able to provide enhanced discovery opportunities for streaming games, mini-apps, mini-games, chatbots, and plug-ins that are found within their apps.

Additionally, mini-apps, mini-games, chatbots, and plug-ins will be able to incorporate Apple’s In-App Purchase system to offer their users paid digital content or services for the first time, such as a subscription for an individual chatbot.

Microsoft must be particularly happy, having recently acquired Activision (after much regulatory back and forth).

Prior to this, cloud gaming wasn’t a thing in the App Store since the services would’ve had to submit every game independently for review. They resorted instead to running their game streaming inside the mobile browser, which as you can imagine is a very sub-par way of doing things.

🇰🇷👮🏻🇨🇳 SK Police Raids Chinese Semiconductor Smuggling Ring

I wrote about this a few times. I’m sure this is just the tip of the tip of the tip of the tip…. of the Iceberg:

The investigation revealed that 96,000 semiconductor IC chips worth approximately 13.9 billion won (US$11.6 million) were smuggled through Company A, with 53,000 of these chips, valued at 11.8 billion won, being classified as strategic items. [...]

These chips, designated as strategic items in 2020, convert analog signals to digital in communication relays and are subject to export and import restrictions

I wonder if these are Texas Instrument chips? Or maybe from Analog Devices?

The IC chips in question, produced by a U.S.-based semiconductor manufacturer, are supplied domestically only through official Korean distributors, which require an end-user certificate and a non-re-exportation commitment from the importer for distribution management. Thus, importing by entities other than the end-user, like Company A, is normally impossible.

Company A circumvented these restrictions by having domestic telecommunications equipment development companies import more IC chips than needed through official Korean distributors, then obtained the excess for smuggling. The chips were repackaged in small quantities and disguised as sample products for shipment to China without obtaining the necessary export permission from the Minister of Trade, Industry, and Energy.

Their profit-making scheme was meticulously planned. They inflated the value of semiconductor components worth only 4 million won to 7.5 billion won when exporting them to Hong Kong, falsely declaring this to customs and submitting fabricated documents to banks to receive the smuggling proceeds. The remainder of the funds was brought into the country through illegal means such as money laundering.

Source.

👜 LVMH Sankey 💍 💰💰💰

This Sankey chart of LVMH by product segment shows just how profitable Fashion & Leather goods are compared to the rest. Look at these margins!

Perfume and cosmetics can also be very profitable, but I think LVMH is a bit sub-scale compared to the biggest players (but it’s not an industry I know very well, so I could be wrong — maybe there’s a temporary situation depressing margins in 2023 and they are usually higher ¯\_(ツ)_/¯ )

---

🧪🔬 Liberty Labs 🧬 🔭

🤖🤝🙋🏻‍♂️ Human Go Players Sharpened by AI

Fascinating stuff:

For many decades, it seemed professional Go players had reached a hard limit on how well it is possible to play. They were not getting better. Decision quality was largely plateaued from 1950 to the mid-2010s:

But then!

Then, in May 2016, DeepMind demonstrated AlphaGo, an AI that could beat the best human Go players. [...]

After a few years, the weakest professional players were better than the strongest players before AI. The strongest players pushed beyond what had been thought possible.

Interestingly, this didn’t happen right after AlphaGo — unlike the 4-minute mile, it wasn’t enough just to know it could be done — but after open-source models became widely available:

the trend shift actually happened 18 months after AlphaGo. This coincides with the release of Leela Zero, an open source Go engine. Being open source Leela Zero allowed Go players to build tools, like Lizzie, that show the AI’s reasoning when picking moves. Also, by giving people direct access, it made it possible to do massive input learning. This is likely what caused the machine-mediated unleash of human creativity.

I can’t help but wonder, where else will this apply?

Or maybe: Where else WON’T this apply? 🤔

Via Henrik Karlsson

‘The World's First Malaria Vaccine Program for Children Starts Now’ 🦟🚫

Malaria is one of humanity’s biggest causes of death and suffering. I hope it goes the way of smallpox someday:

Beginning today, Cameroon, a Central African nation which experiences 2.7 million cases of the disease each year, will start rolling out the world's first routine childhood malaria immunizations using a vaccine called RTS,S or Mosquirix, made by the pharma company GlaxoSmithKline. The vaccine targets sporozoites, the transmissible forms of the malaria parasite, and neutralizes them before they can enter the liver and multiply in their thousands.

With 48 percent of all hospital admissions and 67 percent of childhood deaths in Cameroon linked to malaria, the hope is that this new rollout will help relieve the considerable burden the disease places on the country’s health care system. [...]

The rollout is expected to expand swiftly. Twelve African countries will receive a combined total of 18 million doses of RTS,S over the next two years through Gavi, the vaccine alliance that ensures immunization access in some of the world’s poorest nations.

RTS,S is just a first step, and a better and cheaper vaccine appears to be coming soon too:

The WHO added R21 to a list of prequalified vaccines last month, following studies which showed it reduced symptomatic cases of malaria by 75 percent in the 12 months following a three-dose series. At a cost of $2 and $4 per dose, it’s approximately half the price of RTS,S, and the production capacity is far greater.

Source.

🔌🧠 Musk's Neuralink Achieves First Human Brain Implant

Musk tweeted:

The first human received an implant from @Neuralink yesterday and is recovering well.

Initial results show promising neuron spike detection.

The first @Neuralink product is called Telepathy

Enables control of your phone or computer, and through them almost any device, just by thinking.

Initial users will be those who have lost the use of their limbs. 

Imagine if Stephen Hawking could communicate faster than a speed typist or auctioneer. That is the goal.

Back in September 2023, Neuralink got approval for its first human trials and recruited subjects with certain conditions:

The six-year initial trial, which the Elon Musk-owned company is calling “the PRIME Study,” is intended to test Neuralink tech designed to help those with paralysis control devices. The company is looking for people with quadriplegia due to vertical spinal cord injury or ALS who are over the age of 22 and have a “consistent and reliable caregiver” to be part of the study.

Three different things will be studied during the trial:

The first is the N1 implant, Neuralink’s brain-computer device. The second is the R1 robot, the surgical robot that actually implants the device. The third is the N1 User App, the software that connects to the N1 and translates brain signals into computer actions. Neuralink says it’s planning to test both the safety and efficacy of all three parts of the system.

I certainly hope it works and can help paralyzed people have better lives.

---

🎨 🎭 Liberty Studio 👩‍🎨 🎥

I couldn’t decide between these two images I generated, so here are both ¯\_(ツ)_/¯

🔊🧠 Sound Wave Neurotechnology!

Did you know that you can use various frequencies of sound waves in sequence to make your brain feel all kinds of things. Change your mood, concentration level, creativity, or even subjective experience of life?

Yeah, it’s called music!

Music is perhaps the most powerful and developed form of neurotechnology we have.

Its ability to change our conscious experience with mere sound is extraordinary.

--Eric Wollberg

Neurotechnology, I love that way of putting it:

I really believe in the power of music - and I mean literally the power of musical tones - to rearrange the way you can think.

--Michael Azerrad

h/t Jim O’Shaughnessy (☘️)