529: Anthropic CEO on AI Future, SpaceX Chopsticks, Tesla's Robotaxis & Optimus, Satya Nadella's Secret, Google Goes Nuclear, and Victorian VR

"Like any discussion of the future, it’s largely about the present"

If we are to achieve things never before accomplished we must employ methods never before attempted.

—Francis Bacon

🏛️🧂🧂🧂🧂🧂🧂🫳🌾 I was thinking about the phrase “salting the Earth”, a metaphor so powerful that it implies not just defeating your enemy now, but ensuring that the very soil where they lived is infertile for the future! 💀🪦😵

First, some trivia:

• The famous story of Romans salting Carthage after its destruction in 146 BCE is a modern myth. This tale didn't appear until the 1800s and has no basis in ancient sources. Additionally, Rome was becoming increasingly dependent on foreign grain sources and wouldn't have ruined potentially productive farmland.

But the practice is mentioned in some very old Hittite and Assyrian texts as well as the Book of Judges in the Bible. They not only talk about doing this with salt but also with the seeds of weeds and undesirable species that would make agriculture harder (especially at the time — you can’t buy herbicides at Home Depot in 1720 BCE).

Great story, but what about the practicality and logistics of it? Is it realistic?

I can’t help but think that the amount of salt required to do the job over any area of significance would be *gigantic*. Given that salt was an expensive commodity at the time, it’s unlikely this was done on a large scale.

A bit of research reveals that many historical references to "salting the earth" were likely symbolic or ritualistic, not a practical attempt at destroying fertility.

Plus:

Effectiveness and Practicality

- Amount of Salt Required: To significantly impact soil fertility, an enormous amount of salt would be needed. Modern studies on soil salinization suggest that salt concentrations need to be quite high to severely impact plant growth.

- Cost and Availability: In ancient times, salt was a valuable commodity. Using large quantities to destroy land would have been prohibitively expensive and impractical.

- Natural Processes: Rainfall and natural water cycles would eventually wash away much of the salt, especially in non-arid regions

- Limited Quantities: Given salt's value and scarcity in ancient times, it's unrealistic to expect that large quantities would be used for destroying land.

- Other Purposes: Salt was far more likely to be used for preserving food, as a trade commodity, or even as a form of payment (hence the term "salary")

It’s still a cool expression, though!

🛀💭📺🎥🎞️🍿🎬 I was thinking about how I can watch certain films multiple times and still enjoy them just as much even though I know what’s going to happen.

This leads me to wonder: How much of my enjoyment comes from surprise and suspense?

Is the logical conclusion of this that I could read the plot of a film before seeing it for the first time — fully spoiling it — and still enjoy it as if it were my second time? Or does it only work the second time because I was surprised the first time?

Is it the *memory of the surprise* that makes it work subsequently, or do films that have enough depth to be watched multiple times by definition have more than just the element of surprise going for them?

Maybe great films transcend the need for surprise, offering deeper meaning — or maybe films that effectively achieve suspension of disbelief can immerse us so fully that in the moment we momentarily forget what we know is coming next 🤔

🦃🍁🇨🇦 vs 🦃🇺🇸🏈 I spent the weekend with my parents and sister for Canadian Thanksgiving. It was a fun time. The kids loved poking the campfire with sticks. Seeing the autumn colors is always a good brain massage 🧠💆🏻‍♂️

But it made me wonder why Thanksgiving is celebrated on a different date in Canada and the US. Here’s what I found:

Canadian Thanksgiving

The origins of Canadian Thanksgiving can be traced back to 1578 when English explorer Martin Frobisher held a feast in Newfoundland to give thanks for his safe arrival in the New World2. This celebration predates the American Thanksgiving by 43 years.

So it’s actually the OG!

American Thanksgiving

The first American Thanksgiving is traditionally associated with the 1621 feast shared by the Pilgrims and Native Americans at Plymouth Colony

The main difference, as I understand it, is that it’s a more low-key holiday here than in the US.

I prefer the American tradition, but I prefer the Canadian calendar. Thanksgiving in late November is too close to Xmas and New Year. I like to spread my holidays around a bit more!

---

🏦 💰 Business & Investing 💳 💴

🤖🚀 Anthropic CEO: ‘How AI Could Transform the World for the Better’ 🌈🌎

Dario Amodei, the CEO and co-founder of Anthropic, wrote an essay about what he envisions as the potential benefits for humanity of having access to powerful AI (assuming that the downsides are taken care of — he didn’t want to get into that in this essay).

Anthropic has dedicated a lot of time to downsides in the past, so it’s interesting to see Dario focus on the other side of the coin. However there’s a logic to focusing on risks and problems even if you think the upside is huge:

one of my main reasons for focusing on risks is that they’re the only thing standing between us and what I see as a fundamentally positive future. I think that most people are underestimating just how radical the upside of AI could be, just as I think most people are underestimating how bad the risks could be.

I encourage you to read the essay and make up your own mind:

• 📄 Machines of Loving Grace: How AI Could Transform the World for the Better

I don’t agree with everything Dario says, and the future always is unknowable, but it’s thought-provoking stuff nonetheless.

Here are a few of my highlights:

Avoid “sci-fi” baggage. Although I think most people underestimate the upside of powerful AI, the small community of people who do discuss radical AI futures often does so in an excessively “sci-fi” tone (featuring e.g. uploaded minds, space exploration, or general cyberpunk vibes). I think this causes people to take the claims less seriously, and to imbue them with a sort of unreality.

To be clear, the issue isn’t whether the technologies described are possible or likely (the main essay discusses this in granular detail)—it’s more that the “vibe” connotatively smuggles in a bunch of cultural baggage and unstated assumptions about what kind of future is desirable, how various societal issues will play out, etc. The result often ends up reading like a fantasy for a narrow subculture, while being off-putting to most people.

This is definitely a thing. I’m sure that as a sci-fi nerd I’ve been guilty of it in the past — it’s hard to avoid when you’re very familiar with certain futuristic concepts that the rest of the population sees as really out there and far-fetched.

But an AI that you can have a conversation with seemed like that not so long ago, so I suspect that sci-fi nerds will keep having an advantage when it comes to groking new technologies as they come in.

Biology is probably the area where scientific progress has the greatest potential to directly and unambiguously improve the quality of human life. [...]

Experiments on cells, animals, and even chemical processes are limited by the speed of the physical world: many biological protocols involve culturing bacteria or other cells, or simply waiting for chemical reactions to occur, and this can sometimes take days or even weeks, with no obvious way to speed it up. [...]

If our core hypothesis about AI progress is correct, then the right way to think of AI is not as a method of data analysis, but as a virtual biologist who performs all the tasks biologists do, including designing and running experiments in the real world (by controlling lab robots or simply telling humans which experiments to run – as a Principal Investigator would to their graduate students), inventing new biological methods or measurement techniques, and so on. It is by speeding up the whole research process that AI can truly accelerate biology.

We may get there if we continue making progress on agentic AI that can dump an arbitrary amount of inference into a problem and just “keep trying” until they get somewhere — a bit similarly to OpenAI’s Strawberry “O1” models that use a lot more inference to keep “reasoning” about a problem and then evaluate the various paths taken to try to pick the more successful ones.

To get more specific on where I think acceleration is likely to come from, a surprisingly large fraction of the progress in biology has come from a truly tiny number of discoveries [...] There’s perhaps ~1 of these major discoveries per year and collectively they arguably drive >50% of progress in biology. [...]

I think their rate of discovery could be increased by 10x or more if there were a lot more talented, creative researchers. Or, put another way, I think the returns to intelligence are high for these discoveries, and that everything else in biology and medicine mostly follows from them.

In other words, there are specific points of leverage where intelligence matters most.

There are bottlenecks when it comes to having enough smart researchers and the number of hours that they can work on things. Not to mention that humans often cluster in the most exciting and prestigious fields leaving more low-hanging fruit elsewhere — we could target these first with “AI biologists” and may be able to increase the number of discoveries that move the needle for the whole field (CRISPR, mRNA, genome sequencing, computational protein design, etc).

my basic prediction is that AI-enabled biology and medicine will allow us to compress the progress that human biologists would have achieved over the next 50-100 years into 5-10 years. I’ll refer to this as the “compressed 21st century”: the idea that after powerful AI is developed, we will in a few years make all the progress in biology and medicine that we would have made in the whole 21st century.

Can you imagine!

I’d be particularly interested in seeing progress when it comes to the diseases of aging, since they are the cause of most of the deaths and suffering in the richer countries (and the whole planet is becoming richer, so they’ll be the main problem everywhere eventually).

There’s a section on neuroscience, which I’ll skip in the interest of space. I want to give a couple of highlights on Dario’s economic predictions:

I am not as confident that AI can address inequality and economic growth as I am that it can invent fundamental technologies, because technology has such obvious high returns to intelligence (including the ability to route around complexities and lack of data) whereas the economy involves a lot of constraints from humans, as well as a large dose of intrinsic complexity.

Every time we expect significant productivity gains from new technology, we tend to be disappointed, largely because there’s so much inertia and friction in the system, and diffusion is slow.

I don’t know if powerful AI will diffuse better because it can “slot” itself in the system where humans would be, at least when it comes to white-collar work 🤔 (we’re talking about much smarter and more agentic AI than what we have today, to be clear)

Economic growth. Can the developing world quickly catch up to the developed world, not just in health, but across the board economically? There is some precedent for this: in the final decades of the 20th century, several East Asian economies achieved sustained ~10% annual real GDP growth rates, allowing them to catch up with the developed world. [...]

On the optimistic side, many of the health interventions in the previous bullet point are likely to organically increase economic growth: eradicating AIDS/malaria/parasitic worms would have a transformative effect [...]

There’s plenty more in the essay. Like any discussion of the future, it’s largely about the present. What matters most is how we act *now* to steer things in a positive direction and maximize how great the future is. 🌈🗓️

🚕 SpaceX’s Chopsticks 🥢🚀, Tesla’s Robotaxi & Optimus 🤖

Speaking of AI and robots, there’s been news in the Muskverse lately.

The coolest one was probably from SpaceX, with the successful test of the “chopsticks” to catch a landing Spaceship — an incredible display of mastery of the whole technological stack.

But aside from saying “wow”, I don’t have too much to add there.

Even if remote controlled by a human, it’s still very cool

On Robotaxis and Optimus, I have more to say because these projects are like inkblots: You can see what you want in them.

[➖] If you are negative about Musk, it’s easy to say that this is more smoke and mirrors and empty promises, after years and years of promising that full self-driving was just around the corner; Robots that appear autonomous at first glance but are later revealed to have been remote-controlled by humans; robotaxis that are still far from production, will cost an unknown amount (can we trust Musk’s word on “less than $30,000”? The Cybertruck was claimed to have a base model at “$39,000” — quicky revised to “under $50,000” — and the Model 3 was “$35,000” when first announced), almost no technical details have been revealed, they can only seat two, have scissor car doors ✂️ that are more likely to fail, etc.

[➕] If you are more positive about Musk, you can interpret this as being one more step in the direction of these goals. Things may be taking longer than expected, but it’s because the goals are so ambitious that it was going to take a long time in any case. At least with Musk’s urgency and habit of setting impossible deadlines for his team, we’re progressing faster than through a more conventional process. Waymo may seem ahead now, but if Tesla can get to FSD with significantly fewer expensive sensors (LiDAR), it may have a huge cost advantage.

Optimus may have been remote-controlled, but it’s still a humanoid robot that appears to work well. Getting the hardware to this point was a major challenge, and with advances in AI — including Musk’s xAI’s 100k+ GPU cluster — and the large team of talented engineers working on the software, they’ll get to autonomous robots (at least for certain types of tasks). Possibly before the competition and with a lower cost basis.

Where do I stand on all this? 🤔 I try to keep both sides in mind, but considering Musk’s track record when it comes to this kind of engineering, I lean toward the positive side and expect that he won’t give up until both products are fully realized and actually work.

Now I’m not saying anything about the financial impact of these products, what competition may do (it can be easier to fast-follow than to innovate), and whether the stock is already pricing in these developments. I don’t follow Tesla’s financials closely enough to have an informed opinion.

📞🍸 Satya Nadella: Networker Extraordinaire 🕸️

Interesting piece about Microsoft’s CEO’s networking habits:

Call Nadella Microsoft’s chief information gatherer. A lot of big tech CEOs spend time comparing notes with other leaders in the industry, quizzing customers and hobnobbing with peers. But Nadella has elevated the practice to one of the pillars of his job [...]

Nadella, 57, has instructed his staff to schedule two calls a day with CEOs of other companies, according to someone who works with him. He typically uses part of his weekends to connect in person with executives of other companies in the Seattle area

Two calls a day! Even if he skips some days and most weekends, this could still be 500 calls a year!

His regular interlocutors include Sam Altman, CEO of OpenAI, to which Nadella has committed more than $13 billion. He has also periodically called Srinivas [CEO of Perplexity], Personal AI CEO Suman Kanuganti and Moveworks CEO Bhavin Shah, probing them on how they’re training models and what Microsoft could do to make its Azure cloud services more appealing for small AI developers

I wonder if regulators would allow Microsoft to buy Perplexity if they tried 🤔

He grills venture capitalists in his Rolodex, including Madrona Ventures’ S. Somasegar and Israeli cybersecurity investor Gili Raanan, for intelligence on rising startups, and he regularly speaks to international startup founders like Anthony Tan, CEO of Malaysian e-commerce app Grab.

Everytime Nadella calls, Somasegar says, he asks two questions: What new startups are you excited about, and what new people have you met that would be good to know?

The seeds of many deals are planted this way.

Nadella had been in contact with Mustafa Suleyman for a while before he even founded Inflection, which Microsoft kind of acquired (in a weird acqui-hire + IP licensing structure).

🗣️ Morgan Housel: ‘This Was Never Easy: A Brief History of Nostalgia’

My friend MBI (🇧🇩🇺🇸) recently recommended this one, even mentioned that he listened to it twice, so I *had* to check it out:

• 🎧 This Was Never Easy: A Brief History of Nostalgia

It’s not a new concept to me, but it’s one of these “simple but difficult to keep applying” ones that goes just enough against the way we’re wired that we need to refresh the concept once in a while to avoid falling back into default mental patterns.

---

🧪🔬 Science & Technology 🧬 🔭

Simplified diagram showing the Kairos molten salt reactor.

🔍🤖🔌⚛️ Google Joins the Nuclear Party, Orders Small Modular Reactors

Following in the footsteps of Microsoft, AWS, and Oracle:

Google has ordered six to seven small modular nuclear reactors (SMRs) from Kairos Power [...] with a total capacity of 500MW, helping Kairos, a seven-year-old start-up, to bring its first commercial reactor online by 2030 and additional reactors by 2035. [...]

Kairos, based in Alameda, California, has developed a reactor cooled by molten fluoride salt, rather than water. In December, it received a construction permit from the US Nuclear Regulatory Commission to build a 50MW demonstration reactor in Tennessee called Hermes.

What’s neat about this is that this is a new build, not just buying capacity from an existing plant or refurbishing an old one (which is also great — “Microsoft announced that it would commit to buying 20 years’ supply of electricity from the mothballed US nuclear power plant Three Mile Island if Constellation Energy restarted the site”).

I’m glad to see SMRs getting some love, but the next milestone I’m waiting for is for a large new build to be commissioned by Big Tech. That’ll be quite the signal to the rest of the market!

🧬🏅 2024 Nobel Prize in Chemistry — Untangling Life’s Secrets with Software

This is a particularly fun one for me this year because I’ve been following David Baker’s career for a loooooong time.

I became interested in scientific distributed computing around 20 years ago. I first heard about Seti@home (analyzing radio telescope signals for patterns that might indicate intelligent life out there), but the one that really grabbed my interest was Rosetta@home in 2005. This was a computational protein design software that tried to harness the idle CPU cycles of thousands of online volunteers to crack one of biology’s hardest problems. They were working on building the software and using it to work on actual biomedical and scientific problems.

I remember running this on my computer plus my parents’ computer during the winter (when the waste heat helps keep us warm up here in Canada), as well as Folding@Home sometimes (but Rosetta was always my fave).

I followed all of Baker’s interviews and was part of the Baker Labs forum where the science and software were discussed.

It’s so very cool to see him co-win this almost 20 years later!

And of course, Demis and Jumper’s work spearheading AlphaFold2 is just incredible. Standing on the shoulders of Baker and bringing it to the next level!

(Note how directly connected to Dario’s vision for AI-powered progress this is)

---

🎨 🎭 The Arts & History 👩‍🎨 🎥

👨‍🎨 Portail Paintings aka Victorian Virtual Reality! 🤯

It seems obvious once you see it in action, but I had never thought of how cool the effect could be.

(if you just want to see the effect, skip to 2:15)