359: Qualcomm, Neural Remasters, Google, Cloudflare, NeRFs Special FX, Time Wealth, and Unreal 5.1

"Time is undoubtedly our most precious resource"

A young Bob Dylan aka Robert Allen Zimmerman

👂🏻🎶 🤖🧠 🎧 I bet we’ll start seeing “neural remasters” of music pretty soon.

There’s plenty of great music that has mediocre-to-terrible recording quality, mostly because it used to be a lot harder and more expensive to make something sound good before the modern digital tools era (the difficult period includes the early digital era before the tools became good and engineers really knew how to use them).

It was partly the equipment and studio time, but also the tacit knowledge that was a lot harder to get pre-Internet. Now, thanks to YouTube and audio-nerd forums, you can learn advanced tricks on how to mic drums and use Pro Tools without spending years working in the industry.

What if you trained a machine learning model on a bunch of albums of a certain genre that are considered to have really good sound, then fed it the multi-track recording masters of an album and had it try to improve on the sound and mix?

Purists will say it’s heresy, but I’d be curious to hear some classic old hardcore and metal albums that were recorded in garages on 4-track equipment get a really crisp, punchy sound with actual low-end. I’m sure a bunch of early hip-hop records could also benefit from better production value. In fact, almost anything independent pre-2010 could probably benefit.

I’d never want those to *replace* the original recordings, but it’d be fun to have two versions to pick from, a bit like how a bunch of old Blue Note jazz albums have been remastered by Ruddy Van Gelder, or the early Pearl Jam and Nirvana albums have been re-released with new mixes.

I’d love to hear the early Leonard Cohen or Bob Dylan stuff with a fuller sound.

Or what about a live recording where the performance is great, but the sound kind of sucks? Muddy and lacking bass, full of echo, etc? What if you could train a model on the very-best sounding live shows and try to inject some of that sound into bad recordings?

As long as it’s purely additive to the catalogue and doesn’t replace the originals, I think it would be really interesting.

🚨 🗣🗣🗣🎙📖 In case you missed it, here is a new podcast with two friends-of-the-show, fresh out of the oven:

• 🎧 Claude Shannon fan club with Jimmy Soni and David Senra 📖

👩‍💻🖥🖥 Friend-of-the-show and supporter (💚 🥃) Conor Mac asked:

Looking to upgrade my desk set up. 

If you had to pick ONE thing to level up a set up, what would you recommend?

My advice:

A larger monitor.

Human working-memory is very limited.

Keeping more things visible HELPS THINK BETTER (if you do it right and don't just use the extra space to add distractions).

Having more of what you’re writing visible, having both your writings and research materials, etc.

Diminishing returns apply, I’m not saying you should work on a 65” OLED TV... but I see so many people doing complex mental stuff (research, writing, coding, visual creation) on 12" laptops and it feels like such a low-hanging fruit.

💚 🥃 🐇 This is the free edition of Liberty’s Highlights with 13,400+ subscribers.

You can get 2 extra editions/week full of juicy stuff + access to the private Discord 🗣🗣🗣 community by becoming a paid supporter (it’s quick & easy).

Paid posts since last week 🐇:

• 🔒 357: Cloud Stocks Performance, iPhone Killers, FTC vs Microsoft, Wind Power, LNG, Clarinet Investor, and Hydropower

I like the mini-essay on the multiple waves of “smartphone killers” and why the smartphone endures and is still *such a bargain* when you look at how much utility it packs. The story of the clarinetist who made $170m in the market is pretty wild too.

• 🔒 358: Buffett: Agency vs Determinism, Twitter, Tech Rebalancing, Paul Graham, Google TPUv4, and Nvidia Ray Tracing

I really like the Franz Kafka bit about self-doubt in the intro (that part is free to read if you click). Thre’s also an item about how to best get kids interested in science that is important (change your kid’s life!). I’ll try to apply the lessons with my kids. There’s also a neat video on the making of Kubrick’s Dr. Strangelove.

If you click the links above, you can see the intros for free and there’s a link to get a 7-day free trial of the paid version.

---

🧩 A Word From Our Sponsor: Heyday 🧠

Do you have 100+ browser tabs open right now? 😬

Give your memory a boost with Heyday, the research tool that automatically saves the content you view, and resurfaces it within your existing workflows. 👩‍💻

It’s like cheat codes for your memory. 😲💡

🧩 Give your memory a boost today 🧠

---

🏦 💰 Liberty Capital 💳 💴

Quaaaaaaalcomm!!!!!1

I don’t know what I’m doing with this sub-head. It may be a Wrath of Khan reference 🤔

I loved this episode by friends-of-the-show Ben and David (💚 🥃), in good part because it helped recalibrate my appreciation for Qualcomm.

• 🎧 The Complete History & Strategy of Qualcomm

I got to know the company during a period when a lot of people around me saw them mostly as a kind of leech on the tech ecosystem, extracting huge amounts of money from its armies of lawyers and throve of patents (licensing them as a percentage of products sold, like they deserved a cut of everything else in the product — though I suppose that Apple is getting a taste of its own medicine there, considering its App Store behavior, but that’s another story).

Anyway, I didn’t have the greatest opinion of Qualcomm. A bit like Oracle. While they are very successful businesses, techies mostly didn’t like having to deal with them.

It turns out that the early history of Qualcomm is *really* interesting. It even involves one-degree-of-Kevin-Bacon with none other than Claude Shannon (see the podcast of the first meeting of the official Claude Shannon Fan Club that I just did with David Senra and Jimmy Soni)!

There’s also a big Sam Walton connection, for those playing founder bingo. 👀

Film star and co-inventor of spread spectrum and frequency hopping technology.

And Heidi Lamar fighting the nazis and inventing cutting-edge tech? What a badass she was!

‘This breakdown of Google headcount is both unbelievable and I guess at the same time totally believable’

Can you imagine what the margins on just search would be if they were maximized?

h/t Modest Proposal (🧐)

Cloudflare raising prices for the first time in... *checks notes*… 12 years! ⛅️

A 25% increase after 12 years.

That's not even keeping with inflation *before* the recent inflation spike... And it isn't *really* a price increase anyway, since they added annual plans that are the same price as the old monthly plans.

It's a way to push users to annual plans to get the money upfront. Not a bad idea.

While I'm proud of our pace of innovation, one of the challenges we have is managing the cash flow to fund those investments as quickly as we'd like. We invest up front in building out our network or developing a new feature, but then only get paid monthly by our customers. That, inherently, is a governor on our pace of innovation. We can invest even faster — hire more engineers, deploy more servers — if those customers who know they're going to use us for the next year pay for us up front. We have no shortage of things we know customers want us to build, so by collecting revenue earlier we know we can unlock even faster innovation.

And that's on top of the fact that the "same" plans today include way more stuff than they did 12 years ago, so there's massive deflation in there. Even the free plans today include more stuff than the paid plans did at the time.

The blog post gives a good history of their evolution, from “four and a half” datacenters to 275+, from a couple of products to a whole menu of them today.

🇯🇵 Japan’s baby fizzle 🤰🏻👩🏻‍🍼

Japan's population of more than 125 million has been declining for 14 years and is projected to fall to 86.7 million by 2060. […]

The total of 599,636 Japanese born in January-September was 4.9% below last year’s figure, suggesting the number of births in all of 2022 might fall below last year's record low of 811,000 babies […]

The number of births has been falling since 1973, when it peaked at about 2.1 million. It's projected to fall to 740,000 in 2040.

😬

🇨🇳 China’s Zero-COVID Policy in one graph 🚫🦠

Looking at how they’re still building quarantine camps with tens of thousands of “isolation pods”, I kind of doubt that they’ll loosen up measures quickly. Especially with the current pro-freedom protests…

Via Chartr

---

🧪🔬 Liberty Labs 🧬 🔭

NeRFs: Absolutely insane way to do neural rendering from photos or videos 🤯

Stuff is getting really sci-fi really quickly. I really encourage you to watch the video.

You can dive even deeper with the video below, which looks at the recent papers/algorithms that make this possible:

🎧 Do you listen to podcasts and audiobooks? Watch YouTube videos? Here’s how to ‘create more time’ 🕰 (no magic involved 🪄)

Time is undoubtedly our most precious resource. It is non-renewable and finite.

Depending on how we allocate it, our results in life can vary greatly. Anyone you can think of — a titan of industry, a genius artist, or a lazy couch potato — they all only have 24 hours each day, just like you and I.

Therefore, anything that saves or reclaims time has *great* value (if you use the saved time wisely).

One of the most beneficial things that I have done in recent years is getting used to listening to podcasts, audiobooks, and YouTube videos of talks/presentations/interviews at a speed of more than 1x.

Many of you are probably already doing this and thinking “duh!”, but I want to talk to those that don’t, because I remember what it sounded like before I got used to it.

It sounded like that “speed reading” stuff that has never worked for me, or that has made my comprehension and retention drop so much every time I tried it that on the net it was negative…

When people first hear about it, they often put their podcast player on 2x speed and it sounds ridiculous, like chipmunks, leading them to think, "People are just pretending that it sounds fine to them; this is crazy..." and then they give up.

It reminds me of the first time a friend of mine tried Scotch Whisky (neat).

All he could taste was the alcohol, so he later admitted that he had secretly thought we were BS’ing when we talked about the variety of flavors we could taste. But he stuck with it, got used to 40% ABV, and after a while, the alcohol taste faded away and he started tasting the rainbow.

It just takes a bit of time for your brain to rewire itself.

It’s the same with audio speeds.

The rate at which we think thoughts, or even read, is much faster than the rate at which most people speak, so we’re comfortable ingesting data at that rate. It’s not about pushing things beyond the redline, it’s about bringing audio throughput close to the rate of other faster input methods.

Here’s my recommendation: Start with the smallest increment, and stick to it for many days until it sounds normal to you. In fact, if you go back to regular speed, it should sound like people are drunk.

Then increase the speed by one notch and stick to it for a while. At some point, you will reach a plateau and find the speed at which your brain is comfortable thinking (it can be different for different people — it’s not a competition, if all you can get comfortable with is 1.25x, that’s perfectly fine, you’re still saving tons of time).

I’ve been doing this for years, listening to things at speeds ranging from 1.5x to 2.5x depending on the content. It is very different to listen to a casual conversation with low information density than to a technical talk on a topic one is not familiar with, or to Marc Andreessen who naturally speaks at 1.75x, or to Jocko Willink who speaks at 0.75x.

My preferred podcast player is Overcast.

It has a feature that dynamically shortens the silences between words, so it can give you a 0.25x boost without speeding up the audio. Just this feature alone has saved me over 900 hours since I’ve been using the app (it keeps track at the bottom of the settings page). When combined with the linear speed increase, I’m sure I have saved thousands of hours — or in other words, I’ve listened to thousands of extra hours of interviews and conversations than I would have otherwise.

That’s real value!

YouTube also has a speed option if you click the “gear” icon. Many interviews and presentations can be listened to at 1.5x or 2x without a problem, and over the years, that has no doubt saved me hundreds, if not thousands, of extra hours. The same is true for most audiobook apps.

If you keep doing that for decades, it really adds up.

I know some of you have been skeptical of this for a long time. It may sound like one of these “guru life hacks”, like micro-dosing or something. But give it a try over a few weeks using my method.

You have very little to lose, but the upside could be a big chunk of your life back, or a lot of extra knowledge and entertainment (don’t look down on entertainment — it’s one of the things that makes life worth living!).

---

🎨 🎭 Liberty Studio 👩‍🎨 🎥

🌳 Unreal Engine 5.1: Who knew foliage was this interesting?

If you’re curious about why Unreal 5.0 was such a big deal to begin with, this is a good primer.