146: Nvidia & ARM CEOs Joint Interview, Strangeness Injection, Facebook's 12-Trillion-Parameters Model, Nintendo's Double-Edged Sword, and Intel's Foundry

"I spent way too much time on these 8-bit glasses..."

Narcissism is always enacted at someone else’s expense.

—Stuart Schneiderman

😈 Wouldn't it be cool if two big pseudonymous Twitter/newsletter accounts had a huge ongoing argument in public — I mean old-school rap beef level — on some highly contentious topic, and both made really good points and kept dismantling each others’ arguments in detail. Back-and-forth, back-and-forth…

Over time, factions would form to support them, and these people in turn would argue among themselves, write rebuttals and debunkings, etc.

Then a while later — plot twist! — it would be revealed that both pseudonymous accounts are actually the same person.

I really want to see that happen at least once in my life.

📖 🐉 ⚔️ Am I the only one who does real-time rewrite of the dialogue in my kids’ books as I read to them?

Some of it is so clunky… I have fun trying to improve it and make the story better.

🛀 To me, the best example of compound interest isn’t found in finance, but in technology.

In my short lifetime, my stuff went from a 2400 bauds modem and monochrome ASCII text CRTs connecting to a local BBS that only had 2 phone lines, to streaming 4K HDR on a 55” OLED flatscreen that hangs on the wall like a painting (and there’s still plenty of bandwidth leftover on the wifi in my house for others to stream HD video on their own scifi-like mobile devices)…

We didn’t get there in a big leap forward, but with year-after-year iteration that built on the previous iteration.

🎧 When you listen to a podcast, and you know what the person speaking looks like, do you see an image of them in your head, or do you “picture” a disembodied voice?

Also, if you picture someone, do you picture them in a realistic environment, or kinda floating in space?

I never really thought about it much, but I realize that if I’m listening to, say, Ben Thompson or Bill Brewster, I don’t picture them sitting at a real desk in a real room with clutter and books and an old t-shirt draped over a chair or whatever.

I’m more in the disembodied voice camp, or maybe a kinda of vague, hazy, very faint notion of what the person looks like.

Interesting how the brain works.

🤪 I’m not big into weirdness for weirdness’ sake. It feels like a very adolescent, hipster mindset to me.


I think you have to inject a certain amount of craziness, weirdness, and exploration of strange stuff into your life…

Why is it so important?

Because it’s how you find a lot of the good stuff!

Look at it from the other side (see how I avoided “invert it”? that’s a peek at how writing is behind-the-scenes right there): What’s the opposite of this strange/weird/crazy stuff?

Normal stuff.

Don’t get me wrong, there’s plenty of good stuff that is perfectly normal, but if you stick to that area, don’t expect your results to be very different from other normal people (because if your head is filled by mostly the same things, your thoughts are more likely to be similar too).

The idea isn’t even to whip out that hipster mindset and think you’re “better than” more conventional people because you’ve found good stuff off the beaten path. That’s not why it’s worth doing.

It’s because you can pretty easily find the “normal good” stuff (it’s very visible — they have a big marketing budget and plenty of fans that’ll tell you about it), and combine it with the “strange good” stuff so that “one” + “the other” = more good stuff for you.

You want more good stuff, right?

Then you have to find a way to inject some randomness/exploration/strangeness into your life somehow (maybe this newsletter serves some of that function for you — happy to be of service! *George Patton Salute*)

h/t to friend-of-the-show and supporter (💚 🥃) MI Capital for sparking this idea with his weird gym routines

🛀 🧅 I wonder if the evolutionary pressures that made onions evolve via layers are similar to those that made trees grow via a ring pattern…?

⏱ Follow-up on the time-travel stuff in edition #145:

Thank you for recommending the film ‘About Time’ (2013).

I’ve seen it, and remember really enjoying it, but I hadn’t thought about it in years and had forgotten that aspect of it. I’ve added it to my “to re-watch” list.

If you’ve never seen it, I remember enough about it to be able to recommend it.

💚 🥃 The best thing you can do to support the existence of this project is to become a paid supporter (make sure you’re logged in your Substack account for the button below to take you to the wonderful support options — annual is just 77 cents per edition, and if just one idea helps you make a better decision this year, it’ll pay for itself multiple times over).

The second best thing you can do is to think about who you know who may like this newsletter, and recommend it to them along with a short description of what you think it’s about and why you like it.

💙 Subscribe now 💙

Investing & Business

The Double-Edged Sword of a Strong Culture, Nintendo Edition

I had a nice video call with friend-of-the-show and day 1 supporter (💚 🥃) MI Capital (RIP his Twitter account), and somehow we ended up talking about Nintendo, possibly because we both bought a Switch in the past year, are impressed by Zelda BotW, etc.

It’s hard to talk about the company without touching on the more idiosyncratic and negative aspects, like how restrictive they are online about fans using their games on Youtube and Twitch and such, and we could’ve talked about how they haven’t leaned enough into mobile gaming too…

But what this made me think about is the fact that when a company has a good culture, the good parts carry forward over decades because they have deep roots, some core things are pretty much set in stone, or written in the corporate DNA.

Which is good, but the flipside is the bad parts can be just as hard to change because they also have deep roots, are set in stone, etc

Interview: CEOs of Nvidia & ARM

So far, both Nvidia and ARM have stayed pretty quiet about their merger deal, which is currently making its way through the regulatory anaconda (it’s a long digestive process).

This may have been a mistake, because the narrative has been largely shaped by Nvidia’s competition, who clearly aren’t looking forward to having an even stronger competitor. This matters because while in theory regulation is about following laws and rules, in practice it’s mostly about politics and narrative.

The interview above seems designed to address this and re-take some control of the narrative.

It’s both interesting in what it says directly, but also on a meta-level, to see how they’re thinking about the whole situation, since the interview feels a bit like an infomercial for the deal (but hey, that’s how the world works — you have to sell your ideas if you need other people’s support to make them happen).

They begin by talking about software in general and the new capabilities of the modern AI/ML and internet era.

Here’s my own notes on what Jensen said (paraphrasing a bit):

Jensen: The single largest change is how software is made and deployed. Used to test it as best we can, place it in a shrink-wrapped box, and then we'd update it ever 3-5 years.

Now ML and AI are augmenting the software engineers and helping create "magic" software.

And the entire loop of creating the software, getting data, training the models to make new software… That entire loop is continuous.

Software is creating software. And software is never finished.

Software is always changing, so the world can't have these fixed, static chips, but rather needs things that are software-defined [like what they’ve been doing in networking and security with the DPU, etc]

On the question of “what is the benefit to ARM customers of this deal?”, Jensen had one of his visionary-type answers. Here’s my own notes (paraphrasing a bit):

Jensen: We have two companies that are large and successful, and we would continue to be successful independently. However, the combination is so special because the two of us are complementary. We don't do the same things, we don't do them in the same ways, and not for the same markets.

ARM is a world-class CPU IP company, and the most popular CPU core in the world.

Nvidia is a platform, technology company. We're about the peripherals, the accelerated computing, the software stack, we're a platform company.

So the combination, which is so complementary, is going to bring new innovations, new ideas to the market. Obviously it's going to bring a lot more scale.

The benefit to ARM customers is going to be more IP, better IP, better roadmaps, and hopefully taking ARM to the far reaches of the diversity of computing that is literally growing in every single direction right now. From cloud to edge to IoT, high-performance computing to microprocessors to accelerated computing -- everything.

The breadth of computing today is just gigantic. [...]

One of the things that people I talked to about the deal are excited about is for Nvidia to bring its leadership in artificial intelligence to the ARM ecosystem, which would help it go everywhere. To the edge, to mobile devices, to IoT.

As an aside, I know exactly how long the pandemic has been going on by looking at the length of Jensen Huang's hair (he’s even got a ponytail now!).

The best signal that it's *really* over is probably when he gets a haircut…

‘Wait Times for Chips Hit Record 18 Weeks as Shortage Deepens’

some areas are beginning to catch up with demand. Microcontrollers, small processors that direct functions in everything from cars to washing machines, saw their lead times decrease by more than a week, Rolland wrote. Lead times for analog chips, devices that convert real-world phenomena such as touch and sound into electronic signals, increased, but at a slower pace than previously, according to Susquehanna.

Optoelectronic components, including chips that help convert solar energy into electricity in solar panels, are increasingly hard to come by, according to Susquehanna. Most of the companies listed in the report as struggling to keep up with orders count carmakers as major customers.

Shortages of semiconductors have been felt the most in the automotive industry, which is forecast to lose more than $100 billion in sales of vehicles it can’t make. Other areas have felt the pinch, too, with many electronics makers, including the biggest companies such as Apple Inc., unable to meet all of the demand for their products. (Source)

How to move the Venn circles of Tech/Finance & Environmentalism closer?

I don’t have an answer on this. I don’t even have a good question. But I wish both sides understood the other side better, because they’re both looking at straw versions of each other most of the time.

Somewhere in the Multi-Verse, Yahoo! is such a different beast…

Turner Novak:

Crazy that Yahoo could have bought Google for $1m (now worth $1.7t), FB for $1b ($941b), and used to own almost half of Alibaba ($574b).

Now they’re being vultured for parts by a PE firm with partners that print their emails to read them

And they could’ve sold to Ballmer for like $45bn (prob a lot of it in Microsoft stock now worth multiples of that).

Family Accounts as Viral Marketing

This is a bit similar to Netflix's lax password-sharing stance…

Which I wrote about in edition #108:

On the discussion about Autodesk’s 12 million “non-compliant users” (an euphemism for pirates — aka arr, not A.R.R.), it reminds me of Microsoft, how they have a huge installed base of pirated Windows and Office software, and despite the lost license revenue, it still brought them benefits because it helps with their network effect (file formats), platform value (more potential customers/users for software built on the Windows API), and importantly, it sucks oxygen away from competitors. It’s harder for them to sell their product when they’re competing with both the paid and a free pirated version of the dominant OS/app.

That oxygen-sucking one is prob similar to how Netflix thinks about all the password-sharing. They may not be paying us, but at least they are watching us instead of a competitor, and forming the habit, which increases the chances that they’ll pay later (either because they get tired of being on someone else’s account or because Netflix tightens the screws on the practice).

💙 Subscribe now 💙

Science & Technology

Can a headline be both archaic and futuristic-sounding at the same time?

China's Sichuan Province closes crypto mines, following Xinjiang, Inner Mongolia, and Yunnan; an estimated 90% of China's mining capacity has now been shut down

(this is Techmeme’s headline for this article)

Facebook’s 12-Trillion-Parameters Recommender Deep-Learning AI Model

With the caveat that different ML models do different things, are built differently, and not all parameters are equal… GPT-3’s 175 billion machine learning parameters now look pretty small compared to Facebook’s latest deep-learning recommendation model (DLRM):

We introduce a high-performance scalable software stack based on PyTorch and pair it with the new evolution of Zion platform, namely ZionEX. We demonstrate the capability to train very large DLRMs with up to 12 Trillion parameters and show that we can attain 40× speedup in terms of time to solution over previous systems.

For more details, you can read the paper here. h/t Dylan Patel

Intel’s Foundry Position + Networking’s Deluge

Dylan Patel:

SemiAnalysis believes that Intel will lag 2 years behind TSMC, until at least 2025, in bleeding edge semiconductor manufacturing process technology. This gap could potentially be argued as larger, and there is a high possibility this gap could be extended beyond 2025.

Leading edge semiconductors are a partially a question of scale. One cannot fund the next, more complex process technology unless you have a sizable business to fund it. Intel has continuously lost total share of leading-edge semiconductors to TSMC and Samsung over the last decade as they fumbled the smartphone business and now, as they cede PC and server market share. The losses have accelerated rapidly in the last few years with Apple, Nvidia, AMD, Apple, Amazon, and other in-house silicon efforts. If they continue to lose share in the long run, they may lose the ability to bankroll future process development.  [...]

Rationally, many of these clients still see Intel as a competitor. TSMC champions that they do not compete with any of their customers. It may seem like a no brainer for the competition to continue to flock towards TSMC as they have been over the last decade

There’s interesting stuff about silicon photonics after that, worth checking the whole piece if the space interests you.

#ShroomBoom, Atai Life Sciences Edition

The rules of journalism make it so that every article about this has to start with the fact that Peter Thiel invested in it, because his name is the thing that the most people will recognize and click on…

Atai is the third psychedelic biotech to go public in the U.S., following in the footsteps of MindMed, which went public on the Nasdaq in April, and Founders Fund-backed Compass Pathways, which listed in September [...]

The company, which describes itself as a drug development platform, was set up to acquire, incubate and develop psychedelics and other drugs that can be used to treat depression, anxiety, addiction and other mental health conditions. [...]

In exchange for a majority stake in the drugs and technologies they’re developing, Atai helps the scientists raise money, work with regulators and conduct clinical trials. None of Atai’s drugs have been formally approved by regulators to date.

💙 Subscribe now 💙

The Arts & History

The Office Update (TPS Report Form S2E13)

My wife and I are now a little over the halfway point of season 2, and I’m starting to better understand why so many people like the show. It’s good, but it’s also very comfort-food-like.

I didn’t expect the will-they/won’t-they dynamic of Jim & Pam to be such a central part of it, but they’re so charming that it works. In fact, I think this romance feels a lot closer to reality than what we usually see on TV.

They’re good at conveying what the small moments mean in the early stage of a relationship, or in this case, a supposedly-platonic-relationship-but-hey-we-know-there’s-something-here.

She falls asleep on your shoulder, you have a few too many drinks together and she lets her hair down more than usual, she visits your house for the first time and you hang out and talk about non-work-related stuff, etc. It all feels a lot more like real-life than the more idealized, high-powered romantic stuff that we usually see.

I can dig it.