431: Where Amazon Leads & Lags, Google's Superweapon, Snapchat AI, Shopify + Adyen, SpaceX @ $150bn, Octopus Brains, and Alan Turing

"Plus ça change…"

If people do not believe that mathematics is simple, it is only because they do not realize how complicated life is.

—John von Neumann

Superintelligence 💚 🥃

🧠 It’s impossible to really imagine superintelligence. If we could conceive it fully, we’d be that smart.

However, there’s a thought experiment that I think is a decent intuition pump.

Imagine that you could slow down time for everyone else. Wayyy down. ⏱️

If someone asks you a question, you would have 25 or 50 subjective years to answer it, but to the outside observer, only a few minutes would have passed.

During that time, you would have access to the internet (Google, YouTube, etc), to any books or published papers you want, to software tools to take notes, do math, 3D modeling, physics simulations, architectural drawings, whatever. Anything you don’t know how to use, there’s plenty of time to train yourself to do.

Superintelligence 💚 🥃

You could even have a team of many smart people working with you for the duration, and they all have access to the same resources and tools. They work together and exchange information so that when someone makes a bit of progress, everyone else can build on it.

I think under such conditions, a lot of very difficult questions could be answered.

From the point of view of someone who isn’t impacted by our thought experiment time dilation, it would look like superintelligence.

That’s not to say that this models all that superintelligence could potentially achieve since even a group of smart humans with access to all human knowledge and tools may not be able to do certain things simply because they can’t even *conceive* of them, but I think it’s still a useful way to think about it.

Btw, don’t take the above to mean that I’m claiming to know if and when this type of superintelligence could emerge. I’ve just been thinking about it more because of all the progress in AI lately.

⏳🗓️🕰️ This graph, created by Tim Urban, shows a rough estimate of the amount of in-person time a parent might spend with one of their children:

It’s kind of scary to think about, and whatever the exact numbers turn out to be (ie. some kids leave home later than 19, others sooner, so live in the same city, others on the other side of the world), the general idea is reminding me that time with my kids is precious.

I often tell new parents in my circle of friends that before I had kids, I used to subconsciously think that you had a “baby” for a loooong time. But after I had my boys, I realized that this phase goes by really quickly. Within just one year they go from being helpless to walking around and saying words.

If you zoom out, the first two decades of a child’s life are a bit like that. They may seem like they’ll last forever, but they go by quickly.

I’m not saying it’s all bad! I’m saying that this reality should be taken into account and we shouldn’t sleepwalk through it, both as children and then as parents.

h/t to my friend Jimmy Soni (✍️📚)

---

🏦 💰 Liberty Capital 💳 💴

🛒 Where Amazon Leads and Lags in Retail 🚚📦📦📦

Despite its gigantic scale and mindshare, Amazon still accounts for less than 50% of US e-commerce in most categories.

It’s a good reminder of just how large retail is in general, and e-commerce as a subset of it.

h/t Kourosh

🤖 Snapchat’s ‘My AI’: What are people chatting about?

From my perspective, Snap is a strange company.

They have 750m users, yet they are mostly invisible to me.

I’m sure part of that is because I’m older than the target demographic, but it’s also a relatively self-contained ecosystem. It doesn’t seem to bleed out into the wider web too much (unlike, for example, Twitter or Reddit, which both punch above their weights).

A little while ago, Snap released an AI chatbot called ‘My AI’ (kind of a crappy generic name).

What’s most interesting about it is what people use it for, which gives an indication of how regular people may use AI:

Today we are thrilled that over 150 million people have sent over 10 billion messages to My AI […]

Here are some highlights for the most popular topics:

• Nearly 6 million conversations asking for art and design inspiration

• Approximately 4 million conversations looking for tips on playing guitar and writing songs

• Over 12 million conversations asking for recommendations on skincare, makeup, nail care, fragrance, sunscreen, and other cosmetics

• Nearly 11 million conversations looking for recommendations on hair care

• Over 8 million conversations inquiring about fitness

• Over 16 million conversations about clothing and apparel

• Nearly 5 million conversations looking for info on consumer electronics, another 5+ million asking about mobile apps, and nearly 4 million asking about mobile phones

• Nearly 4 million conversations about footwear

• Over 25 million conversations about pets

• Snapchatters use My AI to help them learn about cars, with 65 million conversations and counting.

• Over 46 million conversations Asking My AI about soccer

• Over 10 million conversations about celebrity and entertainment news

It’s easy to imagine that the most popular topics and verticals will be of interest to advertisers… Snap is already experimenting with “sponsored links” in those conversations.

I’m also sure that all these interactions are extremely valuable to Snap to help train their models. It’s not Meta scale, but it’s a very valuable resource in this world of private datasets and user signals for reinforcement learning…

📺 YouTube: Google’s AI Superweapon 🤖

Speaking of training AI models using user data, Google is in a very advantageous position thanks to Youtube, which is the biggest repository of videos/images (what are videos if not series of images?), audio, and text transcripts on the web.

AI practitioners who compete with Google say the company may gain an edge from owning YouTube, which gives it more complete access to the video data than rivals that scrape the videos.

That’s especially important as AI developers face new obstacles to finding high-quality data on which to train and improve their models.

Major website publishers from Reddit to Stack Exchange to DeviantArt are increasingly blocking developers from downloading data for that purpose. (Source)

The irony is that regulators, which may try to prevent AI labs from scraping publicly available content on the web to protect publishers will likely just entrench the biggest and most powerful companies like Google and Meta by making their own 1st party data more valuable and making it much harder for third parties to compete and train their models.

Plus ça change…

YouTube data could help Google push multi-modal AI further, with input and output being not only text, but also images, audio, and video. It will no doubt also be very useful for training text-to-video generative AI.

🗣️ YouTube AI Dubbing Tool

YouTube is currently testing an interesting new tool:

YouTube is currently testing a new tool that will help creators automatically dub their videos into other languages using AI [...]

Earlier this year, YouTube introduced support for multi-language audio tracks, which allows creators to add dubbing to their new and existing videos [...]

The AI-powered dubbing product transcribes a video for the creator, then translates and produces a dubbed version. Creators can review and edit the transcription before [text-to-voice service] Aloud generates the dub.

🛒 💳 Shopify partners with Adyen for enterprise customers

It looks like the special relationship between Shopify and Stripe may be turning into more of a normal relationship over time…

As part of a multi-year partner strategy, Adyen is supporting Shopify’s expansion into the enterprise segment through diversified commerce solutions for large-scale merchants around the globe. [...]

Adyen’s payment app on the Shopify Payment Partner Platform is expected to be available later this year and enterprise merchants can expect a more flexible approach to their payments experience. 

In the future when enterprise merchants set up the components of their Shopify platform, they will be able to integrate Adyen as a payments provider.

🚀 SpaceX insider sale values company at $150 billion 💰💰

SpaceX is offering to sell insider shares at a price that would raise the closely held company’s valuation to about $150 billion, according to people familiar with the matter. [...]

A $150 billion valuation would compare with $137 billion reported in January when SpaceX raised $750 million from investors. SpaceX has around $5 billion of cash on its balance sheet, the people said.

📰 The Newspaper industry in two charts 🗞️

Via Chartr

---

🧪🔬 Liberty Labs 🧬 🔭

🐙 Did you know that an Octopus has 9 ‘brains’ and 3 hearts? 🧠🧠🧠🧠🧠🧠🧠🧠🧠 🫀🫀🫀

How cool is nature?

Technically, it’s all connected together so it’s the same brain, but it’s highly decentralized:

Octopuses (along with cuttlefish) have the highest brain-to-body mass ratios of all invertebrates; this is greater than that of many vertebrates. [...]

The nervous system is complex, only part of which is localised in its brain, which is contained in a cartilaginous capsule. Two-thirds of an octopus's neurons are in the nerve cords of its arms; these are capable of complex reflex actions without input from the brain. [...]

Unlike vertebrates, the complex motor skills of octopuses are not organised in their brains via internal somatotopic maps of their bodies. The nervous system of cephalopods is the most complex of all invertebrates.

As for the hearts:

Octopuses have a closed circulatory system, in which the blood remains inside blood vessels. Octopuses have three hearts; a systemic or main heart that circulates blood around the body and two branchial or gill hearts that pump it through each of the two gills. The systemic heart is inactive when the animal is swimming and thus it tires quickly and prefers to crawl.

If you want more, here’s an hour-long documentary on squids and octopuses.

🏃‍♂️ Alan Turing was a great runner

TIL that Turing wasn’t only a genius when it came to mathematics, cryptography, logic, computer science, philosophy, and theoretical biology, but he was also quite the athlete.

Alan Turing achieved world-class Marathon standards. His best time of 2 hours, 46 minutes, 3 seconds, was only 11 minutes slower than the winner in the 1948 Olympic Games. In a 1948 cross-country race he finished ahead of Tom Richards who was to win the silver medal in the Olympics. [...]

'I asked him one day why he punished himself so much in training. He told me, " I have such a stressful job that the only way I can get it out of my mind is by running hard." (Source)

h/t Paul Halpern

☀️ Solar is booming in the US… but so is natural gas 🔥

Nearly 4,470 MW of natural gas-fired generation came online in the first four months this year, up from 551 MW in the same period in 2022, FERC said in the report released earlier this month. 

Utility-scale solar capacity increased by 3,409 MW through April this year, up from 3,064 MW in the year-ago period.

New wind capacity fell to 1,967 MW from 5,161 MW in the same periods.

---

🎨 🎭 Liberty Studio 👩‍🎨 🎥

🐻 ‘The Bear’ Season 2 is out 🧑🏻‍🍳🔪

I wrote my review of the first season in Edition #371 (light spoilers).

I’ve only seen the first two episodes of S2, but I’m glad I’m back in this world. So far it’s very good and the anxiety level has been dialed down by maybe 20%, which is nice.

Don’t get me wrong, it’s still a sometimes stressful show, but I really enjoy these characters and the world of this storied Chicago restaurant.

The writing and acting remain strong — so many shows fall off a cliff after S1, glad to see that it isn’t the case so far…

My favorite scenes are still when two characters go on a side-quest, such as when Carm and Sidney discuss the menu, or Carm has a chat with the girl from the grocery store.

I’m looking forward to the rest of the season. I’ll likely be bouncing between this and season 2 of ‘The Great’ for the next few weeks.

Update: My thoughts on episodes 3-4.