404: Edge Paradox, Sonos, Microsoft Security Copilot, Amazon, Alibaba Fragments, Nvidia, Cancer, Writing Arrival, and Chris Cornell's One
"Like many things: it’s complicated and it depends."
There is only one way to happiness and that is to cease worrying about things which are beyond the power of our will.
—Epictetus
🪳💻 This is 404 reporting for duty, defying error expectations!
🔊🎶 I’ve had an Amazon Echo Gen 2 (2017) in my kitchen for years, and while it’s fine, it doesn’t exactly sound great for music.
As I was researching what could replace it (perhaps the Apple Homepod v2?), the newly released Sonos Era 100 popped on my radar. It has some very interesting design and technical choices. If you want to learn more, there’s a feature piece at The Verge and a podcast interview with Sonos’ main sound designer and CEO.
So far, the Era 100 is at the top of my list and I’m also considering the Sonos Arc soundbar for my living room (I’ll reserve my long story about being a former audio geek who’s now coming around to heretical creations like soundbars for another time…).
Part of the appeal is how both would integrate into a multi-room music system and how much focus Sonos puts on music even for its ‘home theater’ products.
Whatever I end up getting, you’ll get my full review, both for music and as a tube that you talk to and that talks back (to be honest, mostly to tell me the weather, set timers, and say dumb things, at least until Alexa/Siri/Google Assistant get some LLM AI injection).
Oh, and I didn’t know, but Sonos is a publicly traded company (Ticker: SONO).
📲 I was reading something about how ‘TikTok’ is just disposable content that basically won’t leave a trace in your mind and that you won’t come back to a year later, etc.
It’s not wrong, but it’s also not right.
Like many things: it’s complicated and it depends.
Too many people assume that everyone’s Tiktok feed is the same. This also applies to YouTube (Shorts), Reddit, Twitter, Instagram (Reels), and all these things. Same principle.
Two individuals can have very different feeds.
Someone may be watching light entertainment for hours, while someone else may be getting some pretty useful and deep stuff from various nerdy/niche accounts, learning about science, technology, philosophy, getting book and music recommendations, learning about new hobbies like woodworking or weightlifting or archery or whatever.
So please, let’s not generalize too much.
The whole point of these user-generated-content algorithmic feeds is that they are customized to each user and aggregate content from a veeeeery long tail of creators of diverse backgrounds, interests, and depths. It’s always funny when some congressman complains about seeing too much suggestive content — maybe it’s because it’s what they keep clicking on and watching..?
But please — pleeeeease — don’t read this as saying that I think everything about Tiktok/Reels/YouTube Shorts is great. Going too far in one direction doesn’t mean we should immediately whip back too far in the other. Let’s just be nuanced about this.
🛀💭📝👀📎🤖 In a future where it’s trivial to co-write everything AI, where next to the “publish” button there will be a “make this better with AI” button, is everyone soon going to sound smarter than they are?
Just as spellcheckers have made us all into spelling bee champs, maybe we'll soon see HyperClippy pop-ups chiming in, 'Hey Bob, your argument kinda sucks! Need me to whip up a stronger one?'
If all of a sudden I start sounding like a cross between Oscar Wilde, Ernest Hemingway, Shakespeare, Mark Twain, Richard P. Feynman, Winston Churchill, Douglas Adams, P.G. Wodehouse, and Albert Einstein, you’ll know why!
From spellcheckers to intelligencecheckers – brace yourselves!
🥣 I randomly noticed a dark pattern on my cereal. We have a few boxes, and once every few weeks my kids get to go nuts and mix all kinds into the same bowl.
While looking at them lined up on my kitchen counter, I noticed something: Every box of cereal had “8 grams of sugar” on the nutritional label, but if you take the time to look at the portion sizes, they’re all different.
Well-intentioned shoppers who quickly scan for “sugar” on every box may not realize that some of them have 2-3x more sugar than others.
🛀💭 Here’s a tip that sounds super obvious but that almost NOBODY does.
Budget an hour on your calendar to think about a certain problem, and then actually do just that for an hour.
It doesn’t even have to be a specific problem you have, it can be just wanting to better understand something.
💚 🥃 🐇 This is the free edition of Liberty’s Highlights with 16,200+ 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 post since last week 🐇:
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. What have you got to lose?
🏦 💰 Liberty Capital 💳 💴
💵 Investing’s Edge Paradox 💸🤪
Dan McMurtrie had a few interesting tweets that I want to share with you:
I think AI is going to cause an explosion of noise and ultimately make markets less efficient to fundamental information. Focus, quiet, and discipline are going to pay more going forward.I am much more uncertain than Dan on this. I think it could go either way…
Does the world become noisier faster than the tools are able to make sense of the world’s information, or does the power of the tools grow so fast that they can handle the increased noise and find the signal more easily than non-AI-augmented agents could in the Before Times? ¯\_(ツ)_/¯
He also posted this one:
Important to note the edge paradox plays still - people who gain informational edges often look at the market and say “wow this is irrational” and then the instant they are done buying/selling the stock, they get frustrated because “the market is being irrational.”This is a great observation, and very true.
I remember feeling this in the past and not noticing the contradiction/irony.
Nothing says the market had to stop being irrational just because I had bought a position… Or even worse: If I’m wrong and the market wasn’t being irrational, then the way things are may be quite stable and last for a long time…
🛒 Everyone’s Year of Efficiency, Amazon Edition 🚚 📦📦📦📦
Amazon in recent months has started warning customers that an item they’re about to buy has been “frequently returned,” suggesting customers check “the product details and customer reviews” before they purchase.
Some may have a cynical take about this, but I think it makes a lot of sense.
If they have that data and they can save some people a bunch of time by having them double-check something and realize that they were making a mistake, why not? I can roll with it, as long as the way they do it isn’t too annoying.
🤖👮🏻🚔💻 🏴☠️ Microsoft Unveils Cybersecurity Copilot
Looks like we’re in the era of AI assistants for everything.
Soon, when you get a haircut, the hairstylist will be wearing Augmented Reality (AR) glasses that will provide instructions and patterns overlaid on top of your head to show how to cut your hair. 🤖🥽 💇🏻♀️ 💇🏻♂️
Anyway, enough sci-fi. Today’s new Copilot is for IT security:
Security Copilot will combine Microsoft’s vast threat intelligence footprint with industry-leading expertise to augment the work of security professionals through an easy-to-use AI assistant. [...]
although attacks have increased 67% over the past five years, the security industry has not been able to hire enough cyberrisk professionals to keep pace. [...] Security Copilot will simplify complexity and amplify the capabilities of security teams by summarizing and making sense of threat intelligence, helping defenders see through the noise of web traffic and identify malicious activity.
In the same way that Github Copilot is helping coders be more productive by automating part of their work, this will allow cybersecurity professionals to cover more ground and hopefully prevent more attacks, or at least figure out that there was a breach and remedy the situation.
Security Copilot will also continually learn and improve to help ensure that security teams are operating with the latest knowledge of attackers, their tactics, techniques and procedures. The product will provide ongoing access to the most advanced OpenAI models to support demanding security tasks and applications. (Source)
One more product based on the OpenAI base model.
It’s impressive how multi-purpose these LLMs can be when you customize and fine-tune them for a specific task.
🇨🇳 Divide & Conquer: Alibaba to be split into 6 companies ✂️
Alibaba said Tuesday it will split its company into six business groups, each with the ability to raise outside funding and go public
Each business group will be managed by its own CEO and board of directors.
Is this the beginning of more big Chinese companies “voluntarily” splitting themselves up?
If being big gives one power, and the government does not want to share power, then making sure that no company gets too big would certainly be a way to do it 🤔
Could Tencent be next? Bytedance? Meituan?
I’m just speculating ¯\_(ツ)_/¯
Here are the ‘40 Thieves’ entities that will emerge from Alibaba:
The business groups will revolve around its strategic priorities. These are the groups:
Cloud Intelligence Group: Alibaba CEO Daniel Zhang will be head of this business which will house the company’s cloud and artificial intelligence activities.
Taobao Tmall Commerce Group: This will cover the company’s online shopping platforms including Taobao and Tmall.
Local Services Group: Yu Yongfu will be CEO and the business will cover Alibaba’s food delivery service Ele.me as well as its mapping.
Cainiao Smart Logistics: Wan Lin will continue as CEO of this business which houses Alibaba’s logistics service.
Global Digital Commerce Group: Jiang Fan will serve as CEO. This unit includes Alibaba’s international e-commerce businesses including AliExpress and Lazada.
Digital Media and Entertainment Group: Fan Luyuan will be CEO of the unit which includes Alibaba’s streaming and movie business.
🔌⚛️ ‘US Govt offers $1.2 bln for distressed, shut nuclear plants’ ⚛️ 🇺🇸💰
It’s been a while since I covered the nuclear power beat, so let’s have a look at some good news:
The Biden administration said on Thursday it is offering $1.2 billion in aid to extend the life of distressed nuclear power plants which, for the first time, could offer funding to a plant that has recently closed. [...]
about a dozen U.S. reactors have closed since 2013, leaving 92 across the country.
Good move.
These precious, hard-to-replace assets, many of which were victims of a combination of ratcheting regulatory costs + a period during which natural gas was incredibly cheap because of the fracking boom, can still pump out reliable clean power 24/7, help clean up the US power grid, and make it more reliable by providing baseload power.
The funding comes from the $6 billion Civil Nuclear Credit program, created by the 2021 infrastructure law, and will be distributed by the Department of Energy (DOE).
In this second round of program funding, the money is available to plants at risk of closure within a few years, but also for the first time, plants that have stopped operating after Nov. 15, 2021. (Source)
This last bit means that the Palisades plant in Michigan should be able to apply. It closed in May 2022. Hopefully, this brings it back to life.
🇨🇳⚡️ Solar & Wind projects in China facing bottlenecks
It’s not just in the US that wind & solar are facing all kinds of challenges to get their electrons to the grid:
Last year, China installed 152 gigawatts (GW) of renewable capacity, or 76.2% of all new energy capacity [...]
However, its rapid growth has put it into competition with other industries, particularly agriculture, for land use, making it increasingly tough to win land rights for installing solar panels. [...]
China needs to remove the hurdles to construction and promote integration with the grid to ensure its first batch of large-scale wind and solar power bases are put into operation on time, the NEA said in a report following a recent meeting on the issues.
President Xi Jinping has pledged to bring China's total wind and solar capacity to at least 1,200 GW and to cap its carbon emissions by 2030.
The country is also planning to build 450 GW of solar and wind power generation capacity in the Gobi and other desert regions. (Source)
Building transmission lines to remote areas and obtaining land rights to build these vast wind & solar farms does highlight one of the big drawbacks of low-power-density sources: They take a lot of land, and require very long transmission lines since the areas that are sunniest and windiest are not necessarily the same where population centers are.
This is reason #675 why conservationists and environmentalists should be pushing for nuclear power. With a tiny tiny tiny amount of land area, we could produce enough 24/7 clean power for the whole grid.
Think of all the embedded energy (ie. burning diesel fuel, mostly) in all the materials that need to be mined to build these renewable megaprojects?
🧪🔬 Liberty Labs 🧬 🔭
Nvidia’s new AutoDMP AI tool optimizes the macro layout of semiconductor chips up to 30x faster 🤖🐜
Most modern digital chips integrate large numbers of macros in the form of memory blocks or analog blocks, like clock generators. These macros are often much larger than standard cells, which are the fundamental building blocks of digital designs.
Macro placement has a tremendous impact on the landscape of the chip, directly affecting many design metrics, such as area and power consumption. Thus, improving the placement of these macros is critical to optimizing each chip’s performance and efficiency.
Because there are so many variables, it’s very difficult to optimize the whole layout manually, or even with traditional tools.
The [reinforcement learning] method formulates the macro placement problem as a game. The actions of the game agent correspond to possible locations for the macros. By sampling many placement examples, the agent learns to optimize its policy (usually represented by a neural network) to improve the macro placement quality. This quality is estimated based on surrogate objectives such as wire length, congestion, and density. [...]
DREAMPlace computes both wire length and density gradients numerically using GPU-accelerated algorithms enabled by the PyTorch framework. It achieves over 30x speedup on global placement alone. Further work also accelerated precise placement with GPUs, achieving over 16x speedup on industrial benchmarks over CPU implementations.
30x and 16x! That’s quite the improvement — we’re not talking about something that is marginally faster than the alternatives…
Progress on Cancer in one graph 🩻🧫🔬👩🔬
🏋️♂️ Peter Attia’s book is out! 🏃♂️ ⏳
I haven’t read it yet, but I suspect that I’ll find it quite interesting, and you may too, so I thought I’d mention it briefly. Once I’ve read it I’ll share my review.
Peter recorded a podcast about the creation of the book (and audiobook! Some great anecdotes on that) with his co-author here. Learning about how they spent 6 years working on this and how the book died in 2020 made me want to read it even more:
🎨 🎭 Liberty Studio 👩🎨 🎥
Chris Cornell’s cover of U2’s ‘One’ with lyrics to Metallica’s ‘One’ 😲🤯🤘
I randomly saw this comment on Reddit and I *had* to check it out:
On a related note, I recently discovered Chris Cornell's cover of "One".
He intended to cover the U2 song, but when he Googled the lyrics, the first song that came up was Metallica's "One". He decided those lyrics were more interesting and mashed up Metallica's lyrics with the melody from U2, and the result is amazing.👽🐙 How Ted Chiang’s Story of the Your Life was adapted into the film Arrival 📕 → 🎬
I love great art, and I love learning more about how it was created.
What stood out to me watching this is just how much creation is about problem-solving. And the anecdotes about Denis Villeneuve and how he works with people just confirm to me that I love the guy. I can’t wait for Dune: Part 2.
Cancer progress graph is.... WONDERFUL! 💐🌷😍