Plugging in sovereign AI🔌 in service to nature🌿 "heart skills"💝 torso and neoclones💪 the tech coup🏛️ generative infinite games 🎮∞ Wayback is back💾 act one🎬 system zero thinking💭 #2024.43
Plugging in sovereign AI🔌 in service to nature🌿 "heart skills"💝 torso and neoclones💪 the tech coup🏛️ generative infinite games 🎮∞ Wayback is back💾 act one🎬 system zero thinking💭 #2024.43
Show me the gigawatts⚡
Another week flies by… Welcome to *yet another* Memia scan across the latest AI, emerging tech and our exponentially accelerating future(s). The curves keep on going up, an even bigger missive than usual this week — strap in…
ℹ️PSA: Memia sends *very long emails* (especially this week!), best viewed online or in the Substack app.
Listen to this week’s newsletter as an AI-generated podcast:
🗞️Weekly roundup
The most clicked link in last week’s newsletter (other than the NotebookLM deep-dive podcast of the same…these are still popular…) was Callaghan Innovation’s very early POC implementation of a New Zealand government chatbot: GovGPT.
Also, following up last week’s $GOATed story: Marc Andreessen and Ben Horowitz discuss the Truth Terminal saga so far… How An AI Bot Became a Crypto Millionaire. (They’ve sure got the Dr Evil look at A16Z…)
↩️ICYMI
I had a productive Sunday dialogue with my new co-thinking bestie claude-3-5-sonnet-20241022, exploring a more appealing pathway to the future than blunt techno-accelerationism:
Lots of engagement on that post… a rich seam to explore deeper in the next few months…
🌿In service to nature
Starting with three stories about the environment for a change.
💰🌳Toha for Te Kautuku
As the UN COP16 Biodiversity Summit in Colombia wraps up later this week, here in Aotearoa there was an announcement from startup Toha Network. (I first covered Toha way back in Memia 2021.08 and more recently discussed in Memia 2024.25 and Memia 2024.33).
Toha is developing what it calls Digital Public Infrastructure: a protocol (software) which manages payments for verifiable claim-based actions for restoring and regenerate natural ecosystems. At the core is the MAHI digital token, representing:
“a unit of work in service to nature”
Yesterday Toha announced its first pilot project to raise NZ$2.5M to fund the regeneration of Te Kautuku, a 930ha Māori land block on the far East Coast of Te Ika-a-Māui / North Island:
“…Its natural ecosystems range from lowland forests to dunelands, with wetlands and river systems connecting directly to the Pacific Ocean.
… Its remnant forest is a living example of what is possible by protecting and enhancing indigenous biodiversity.
The model being piloted at Te Kautuku has the potential to unlock a new revenue and investment for over 1.4 million hectares of whenua Māori in Aotearoa and indigenous land globally wanting to regenerate their ecosystems.”
The Te Kautuku pilot is the first proof point for Toha’s digital co-investment model for regenerating indigenous biodiversity.
Although early days and operating at tiny scale, Toha remains the most sophisticated regeneration marketplace effort I’ve seen to date — with potential to create a whole new level of functionality to fix the urgent challenge of market pricing for biodiversity.
Unfortunately, the involvement of national airline Air New Zealand as an anchor investor on this first pilot opens the scheme to accusations of greenwashing…. high profile environmentalist Mike Joy doesn’t mince his words (as usual, thankfully):
Perhaps the outcomes will prove Mike wrong… certainly there’s no shame in trying every intervention to try to fix the techno-industrial system… while living inside it. Wishing the Toha team well for their first pilot and beyond! 🫡
🌊The Challenge
Staying with the environment: research from Edinburgh University scientists proposes that halting ocean acidification from pollution could kickstart natural phytoplankton-based carbon sequestration back to previous levels…
“The ocean's surface microlayer (SML) contains much of the planet’s vital phytoplankton — but also attracts toxic forever, lipophilic chemicals, microplastics and hydrophobic black carbon soot from the incomplete combustion of fossil fuels. Concentrations of toxic chemicals are 500 times higher in this SML layer than in the underlying water… [impacting] phytoplankton photosynthesis which may have declined by as much as 50% since the 1950s.”
Furthermore, the principle author of the paper, Howard Dryden, argues that the simplest way of achieving pollution-free oceans is to radically change how we treat wastewater before it flows into them… in particular by using a series of natural settlement lagoons rather than energy-intensive wastewater treatment plants.
More explanation in this 7-minute video from UK independent journalist Xan Phillips: (not sure on the consensus on the science but intuitively feels like a common sense win/win for designing water treatment plants from here on…)
🧹The Cleanup
Speaking of pollution-free oceans…. the Ocean Cleanup project continues to execute and scale… doubling the amount of plastic removed from the Ocean compared with this time last year, with plans to double again in 2025:
💸Monetary policy without QE
In other news… Bitcoin went above US$70K for the first time since it hit a record price of $73,844 in March… a new record looks likely, touching US$73,500 just now…
Even allowing for the *irrational exuberance* of crypto markets, see below my quick sketch juxtaposing two charts: (1) Monetary policy WITH Quantitative Easing (top) and (2) Monetary policy *WITHOUT* QE…
(The US Presidency could well be a poisoned chalice for whoever wins it…)
📈The week in AI
OK, on to the week's AI news and releases - foot back on the gas!
🔌Plugging in Sovereign AI (AKA “Show me the gigawatts”⚡)
A flurry of announcements and reports in the last week which fill in a picture of how frontier AI is now being treated as a hot national security issue in the US… with sudden implications for energy infrastructure and IP protection:
The Biden administration released a National Security Memorandum (NSM) on AI, which focuses on maintaining US leadership in frontier AI systems. The NSM contains policy directives covering:
Accelerating AI adoption across national security agencies
Creating an AI National Security Coordination Group
Expanding AI-enabling infrastructure and energy capacity
Attracting global AI talent through immigration policies
Enhancing counterintelligence efforts to protect AI assets
Collaborating with international partners on AI governance
All of this implies that The Pentagon has fully bought into the “AI Entente” race-against-China polemic articulated by Leopold Aschenbrenner earlier this year in his seminal essay series Situational Awareness - The Decade Ahead. (Worth going back and re-reading that with the new NSM now top of mind…)
Greg Allen and Isaac Goldston of Washington thinktank CSIS recorded an accessible podcast: The Biden Administration’s National Security Memorandum on AI Explained which breaks down the NSM — and speculating what next week’s US elections could mean for its implementation. Basically it’s “foot on the gas” and “foot off the [governance and AI safety] brakes”… until Trump gets elected. Then who knows…
Andrew Curran also charts the rapid emergence of the term “Sovereign AI”, a term used by Oracle’s Larry Ellison and also on a recent Nvidia earnings call by Jensen Huang, no less.
OpenAI has been pushing the GDP and job-creation benefits of a massive AI infrastructure buildout:
As reported in The Hill:
The Hill via @morqon The most comprehensive analysis so far arrived this week in a newly released report from Institute For Progress (IFP) analysts Tim Fist and Arnab Dutta: How to Build the Future of AI in the United States.
It’s a huge report, but summarising the main thesis In four charts:
If scaling “laws” continue and frontier models continue to suck up all available GPUs and TPUs, there could be over 100 GW of new data centre power demand globally by 2030, up from ~45GW currently:
Currently US installed generation capacity sits at under 1,500 GW… with new (mostly intermittent) capacity “In the Queue” for build and connection to the grid … 4-10 year lead times for new grid-connected power in the US is just too long, presenting 2 alternative choices…
…Build it in Brazil or the Middle East…
(…which would mean that the “security” piece of “national security” could be more easily compromised and US$5Bn-training-run models could go walkies…)
…OR build dedicated generation on-site in the US at non-commercial cost rates:
Overall, the IFP report is a detailed, data-rich synthesis of the current race condition predicament facing the US AI industry and government — BUILD… or face growing AI competition from China (and maybe others…)(The report is long but brilliantly human-readable… nonetheless dive in with NotebookLM or Claude and see what you find…)
Finally, despite its intermittent power characteristics, many commentators think it’s likely going to be faster and more economic to build out solar than nuclear:
China has installed a cumulative 770GW (nameplate) of solar generation, with 160 GW added in just the first three quarters of this year.
In Saudi Arabia… the latest tender to build a 3.7 GW solar farm attracted a lowest bid of US$0.0129/kWh.
Casey Handmer, big-thinking physicist and CEO of Terraform Labs explored the land use implications of Solar+Battery to meet AI demand earlier this year: How to Feed The AIs:
“What is this going to look like at scale?
A 1 GW data center (containing roughly a million H100s!) would have a substantial footprint of 20,000 acres, almost all of that solar panels. The batteries for storage and data center itself would occupy only a few of those acres. This is in some sense analogous to a relatively compact city surrounded by extensive farmland to produce its food.“
(See Mind Expanding later on for his latest plan: to flood Nevada…)
🚀At Scale
Alex Wang is the youthful-looking CEO of Scale.ai (which is a “data foundry” providing the data for training massive AI models). At a recent Cisco AI Summit event he had some rich insights on where the frontier of AI is heading…
(Full interview below, from 1:26:00 in):
Using Claude to summarise his most interesting points:
There are five key components of AI Development:
Compute (handled by companies like NVIDIA)
Algorithms (developed by companies like OpenAI, Google, Anthropic)
Data (Scale's focus)
Low-latency networking (emerging as fourth dimension)
Security/AI safety (emerging as fifth dimension)
Around US$200 billion being invested in training next-generation AI models
This is compared to US defence budget of US$800-900 billion
Current monetisation (e.g., ChatGPT) is generating billions, but most use cases not yet realised
Major Challenges in AI Development:
a) Hitting the “Data Wall”:
Running out of publicly available training data - the whole internet is in the training set
Need for more sophisticated "frontier data" - will be a mix of synthetic and enterprise private data
Enterprise private data is starting to become crucial (e.g., JP Morgan has 150 petabytes)
b) Evaluation Problems:
Current benchmarks are saturated
Difficulty in measuring AI system capabilities
c) Chips and Energy:
20 gigawatts of energy will be needed in the next 2-3 years to power new AI data centers - 5 Chicagos worth of energy (cf. IFP report above…!)
Heavy reliance on Taiwan for high-end chip manufacturing
Growing interest in nuclear power production
🤖Scarier than Terminator
On the other side of the optimism ledger, legendary film director, Aotearoa citizen and newly-minted director of Stability.ai James Cameron had some cheery words for a recent conference keynote:
“…AGI will not emerge from a government-funded program. It will emerge from one of the tech giants currently funding this multi-billion dollar research. So then you'll be living in a world that you didn't agree to, didn't vote for, that you are co-inhabiting with a super-intelligent alien species that answers to the goals and rules of a corporation.
An entity which has access to the communications, beliefs, everything you ever said, and the whereabouts of every person in the country via your personal data. Surveillance capitalism can toggle pretty quickly into digital totalitarianism. At best, these tech giants become the self-appointed arbiters of human good - which is the fox guarding the henhouse. They would never ever think of using that power against us and strip mining us for our last drop of cash.
That's a scarier scenario than what I presented in The Terminator 40 years ago, if for no other reason than it's no longer science fiction - it's happening.“
🤥Impossible to definitively know
You have to be grudgingly impressed with Meta’s written response to Australia's Senate Select Committee on AI Adoption questions about the scope, sources, and legality of data used to train the company’s AI models.
15 pages of dense text, monolithic paragraphs… and not a clear answer in sight. For example when addressing whether their models were trained on copyrighted data:
Anyone gearing up for an AI copyright battle may want to reconsider their life choices if they want to go up against this for the next 5 years…
💝“Heart skills”
A recent AI report from Citibank contained, er, mixed messages:
AGI by 2029. ASI in the 2030s…
…but don’t worry… work on your “heart skills” and you’ll be fine:
(Given that deepfakes are now well on the way to uncannily impersonating all human emotional communications and as a result we are all highly vulnerable to limbic-system-hacking by AI, personally my advice would be diametrically opposite: put as many AI self-defence systems as possible between the internet and your cognitive inputs; absolutely *don’t* develop more “heart skills” to open yourself up to emotional hijacking by superintelligent AI!!!)
Benjamin Todd has a pessimistic outlook on what this means for society globally:
@AISafetyMemes exquisitely captures the dissonance with normie land:
📱Other AI industry news
Denmark’s king was joined by Nvidia CEO Jensen Huang in a ludicrous cable-joining ceremony to launch the country’s “first AI supercomputer”, Gefion, built on the Nvidia DGX SuperPOD platform and using 1,528 Nvidia H100 GPUs. (Pretty small compared to the million-GPU data centres being talked about in the US…)
Novo Nordisk via Tom’s Hardware Tesla shares closed up a record 22% in one day, adding more than $150bn to its market value, following strong earnings in their latest quarterly announcement. Volatile. (Just in time for Elon Musk to pay down some of his Twitter acquisition debt…?!)
Character.AI faces a lawsuit after a US teen became addicted to the chatbot and committed suicide - Zvi Mowshowitz looks deeper and finds not a lot to see here…
Another one out the door: Miles Brundage, Senior Advisor for AGI Readiness at OpenAI quit after 6 years tenure (which makes him an OpenAI veteran… we met several years ago at a Partnership on AI conference in Berlin). He leaves with this parting shot:
“So how are OpenAI and the world doing on AGI readiness?
In short, neither OpenAI nor any other frontier lab is ready, and the world is also not ready. “
On which subject….GPT-Next by December? Next month marks exactly 2 years since ChatGPT was first released. Industry rumours have OpenAI is planning to release its next major model codenamed “Orion” which is “100 times more computational volume than GPT-4” to coincide with the anniversary. (Although likely only available to OpenAI enterprise customers initially rather than rolled out to all ChatGPT customers. (This is how it will go from now…)
@Sama enjoying shitposting as always:
📚AI research
I’m about a year behind on this one… Plotting Progress in AI…the evolution of AI benchmarking. (Basically: be very careful saying that “AI will never be able to do X”).
“Explainer: How were these numbers calculated?
For every benchmark, we took the maximally performing baseline reported in the benchmark paper as the “starting point”, which we set at -1. The human performance number is set at 0. In sum, for every result X, we scale it as (X-Human)/Abs(StartingPoint-Human).”Comparing AI risk assessment frameworks in Australia and Aotearoa
∞ 🎮New from Google, Unbounded: research on a new class of interactive experiences they call generative infinite games, essentially video games where the game mechanics and graphics are fully subsumed by generative models
For those interested in going down into the technical details, a thread from Anthropic’s Interpretability team on their recent research, including Crosscoders - which enables a “diff” between pretrained and fine-tuned models:
Anthropic Kaggle Gemini Long Context contest:
With large AI model context windows, methods like vector databases and RAG (that were built to overcome short context windows) become less important, and more direct methods such as in-context retrieval become viable instead.
Gemini 1.5 comes with a massive 2 million token context window (equivalent to being able to remember roughly 100,000 lines of code, 10 years of text messages, or 16 average English novels).
What would you do with such a large context window? Google are running a contest on Kaggle to find out with US$100,000 in prizes.
🆕AI releases
These AI labs they keep on shipping…
Claude can now write and run code to perform calculations and analyze data from CSVs
Microsoft Omniparser is an open-source UI parsing/understanding model that outperforms GPT4V, released under MIT licence:
IBM (yes, IBM!) open-sourced Bee Agent framework to build, deploy and serve agentic workflows at scale.
🎬 Runway Act One introduces some quite exciting possibilities for acting in your own AI-generated movies… (when it arrives…)
Similarly for open source GAGAvatar:
Stability AI released Stable Diffusion 3.5:
Includes multiple model variants, including Stable Diffusion 3.5 Large and Stable Diffusion 3.5 Large Turbo. (Stable Diffusion 3.5 Medium will be released on October 29th).
Highly customizable for their size and run on consumer hardware.
“Open release” free for both commercial and non-commercial use under the permissive Stability AI Community License.
Impressive image quality, with text:
Stability.ai
Ideogram Canvas:
“an infinite creative board for organizing, generating, editing, and combining images.”
Bolt Build any web UI from only a prompt or an image (try screenshotting your favourite website and see what it comes up with…)
🔮[Weak] signals
Non-AI tech signals from near and far futures... pretty rapid scan through.
💾Wayback is Back
The Internet Archive websites http://archive.org, http://openlibrary.org as well as http://web.archive.org (Wayback Machine) came back online as a read-only service after cyberattacks.
(It’s an amazing non-profit service which I’m really happy is able to keep going on its shoestring budget… if you’re so inclined, you can donate here)
⛏️US lithium deposits
The US government pushed through approval of Ioneer's Rhyolite Ridge lithium-boron mine in Nevada — the mine could produce enough lithium to power about 370,000 electric vehicles annually.
Meanwhile over to the East: a recent USGS study using machine learning to map lithium concentrations estimates that the Smackover Formation in Southern Arkansas may contain 5.1 to 19 million tons of lithium dissolved in brine — potentially representing 136% of the United States' current lithium resource estimate.
The familiar cycle continues…
🛰️One less satellite
The Boeing-manufactured Intelsat satellite IS-33e disintegrated in orbit, causing service disruptions across Europe, Africa, and parts of Asia-Pacific to Intelsat customers. The US Space Forces are tracking approximately 20 pieces of the satellite. Not Boeing’s year...
🦿Robots
EngineAI appears to have solved the problem of walking with a human-like gait:
💀Meanwhile the new Torso from Clone Robotics is giving off real Westworld vibes…
As previously covered, Clone is taking a completely different approach to building robots than almost all other firms, developing pneumatic muscles rather than servo motors. As a result, the biomimicry is far closer. Here’s Clone’s remotely operated hand catching a ball:
💪As for their roadmap… start planning for the Neoclones:
But….why the obsession with humanoid robot form factors, anyway?
🎮Drones
Military drones in the field Israel released drone footage of the last moments of the injured Hamas leader Yahya Sinwar before he was killed, attempting to throw a stick to disable the drone. This tech is now being deployed more and more in the field... the future of warfare won’t involve human soldiers.
Indoor drones in the dorm more wholesome: a low-cost (only US$20), open-source motion capture system for tracking and flying drones autonomously, with millimeter-level precision at dorm room-scale, developed by University of Cambridge Computer Science undergrad Joshua Bird:
👓AR/VR
Lifeskin Impressive concept video for switchable AR environments in your own home… entertaining, but… really?
🔗Crypto
What else is happening in Crypto other than Bitcoin breaking through US$73K…?
Web3 IBAN Gnosis Pay wallet (featured a few weeks ago in Memia 2024.40) can now be hooked up to any bank account on the IBAN network in 93 countries:
💨CO2 capturing
(Constantly scanning for a miracle “wonder material” which can be made abundantly and removes CO2 from the air.)
Spotted in Nature: scientists have synthesised COF-999, a new porous, crystalline covalent organic framework (COF) which can capture CO2 from open air:
“COF-999 has a capacity of 0.96 mmol g–1 under dry conditions and 2.05 mmol g–1 under 50% relative humidity, both from 400 ppm CO2. This COF was tested for more than 100 adsorption–desorption cycles in the open air of Berkeley, California, and found to fully retain its performance.”
(The energy required for the cycles could be powered by solar…)
🧀Switzerland in space (one day)
A group of Swiss students achieved a significant milestone in European space technology: successfully conducting the continent's first rocket hop test. The group’s 2.5-metre reusable rocket demonstrator, called Colibri, performed a 105-metre free flight, lasting 60 seconds, with a successful vertical landing, on a budget of under 250kChF(US$287K). AND with a patriotic slice of Swiss cheese attached to the rocket, no less!
(Video: @GruyereSpace). Love this comment:
♻️Recyclable renewable
This is cool - a Tiny Home made from a recycled wind turbine tower:
🧠Mind Expanding
A few links to tickle the neurons (and GPUs…)
💭External (“System 0”) thinking
Psychologist Daniel Kahnemann (RIP this year) famously introduced the concepts of “System 1” (fast, intuitive) and “System 2” (slow, analytical) thinking in his seminal book Thinking, Fast and Slow.
Now a team of Italian neuropsychology researchers have published The case for human-AI interaction as System 0 thinking, proposing "System 0" as a new cognitive model integrating human-AI interaction, operating alongside alongside human System 1 and System 2 thinking. System 0 leverages AI's massive data processing capabilities to enhance human cognition externally…. but at the risk of overdependence and blind trust.
This definitely aligns strongly with how I perceiving my “co-thinking” buddy AIs now and I’m constantly on the lookout for smoother augmentation and integration workflows… I wrote about a similar topic way back in 2017 discussing the earliest BCIs: Why Neuralink and Kernel are trying to solve the right problem at the wrong time, how human brains augmented with AI (in data centres) will organise themselves (“HIML” = “Human Intelligence Markup Language”… which as I guessed at the time is basically modern natural languages…)
…a few years later becomes…
🏜️Terraforming with desalination
BIG thinker Casey Handmer returns, this time proposing Terraforming the American West, re-routing major rivers and pumping desalinated seawater into the desert:
“I think solar plus matched desal can get below 10c/m^3, cheap enough that we can put 5 million acre feet into Nevada, terraforming about 1500 miles of valleys and waterways in that state (>90% remains unchanged though) and creating space for tens of millions of people to find wealth and opportunity in the process.”
(An intriguing prospect…but I wonder has he heard of the Aral Sea, what happens when big river diversion projects have unplanned consequences.)
🌐Nostr
Gordon Brander argues how the fast-growing Nostr protocol's relatively simple relay system offers a balance between scalability and user autonomy — presenting a potential solution to the centralisation problems inherent in other network architectures.
🏛️The Tech Coup
Framelab’s Gil Duran interviews Marietje Schaake about her new book "The Tech Coup: How to Save Democracy from Silicon Valley":
⏳Zeitgeist
Once around the world in the corners I poke around in...
Election watch
Discuss:
“Democracy is the worst form of government, except for all the others that have been tried.” —Winston Churchill
The biggest ever year for Democracy continues…
Japan is the latest country to undergo a snap election, resulted in a surprise loss for the long-dominant LDP. Political instability beckons…
(In case you’ve been living under a virtual rock….) the US *finally* goes to the polls next week. I think everyone is *OVER IT*. Latest exhibits on how broken the US electoral system is:
Between them Kamala Harris and Donald Trump have raised over US$2 Billion of campaign financing. With strings attached…
Elon Musk, one of the world’s richest billionaires, openly offering to pay registered voters in swing states at least US$47 to sign a petition “pledging support” for freedom of speech and the right to bear arms — and the chance to win a daily prize draw to win $1m. Now votebuying out in public rather than disguised…
Meanwhile, climate change is 21st out of 22nd of issues important to US Registered Voters:
Eli Dourado runs a series of Monte Carlo simulations of voting outcomes and examines what happens in the 1.2% probability event of an exact draw in electoral college votes.
🌡️Climate extremes
On to more important matters. Climate Extremes: At the Abyss? is a new documentary film from OoS Pictures EHF. 52 educational minutes well spent:
Are we at risk of crossing planetary tipping points? Are we at risk of pushing the planet toward a trajectory where it could unstoppably drift away from a state that can support life as know it?
Climate Extremes: At the Abyss? explores these questions, presenting expert perspectives and ongoing scientific research on the rapidly changing Earth system and the rising risk of triggering rapid, non-linear, and even self-reinforcing "vicious cycle" feedbacks in critical planetary systems like the Antarctic and Greenland ice sheets, ocean currents (including the Atlantic Meridional Overturning Circulation, or AMOC), and the Amazon Rainforest.
Featuring insights from scientists including Johan Rockström, Daniel Swain, Stefan Rahmstorf, and Samantha Burgess from the Potsdam Institute for Climate Impact Research, the European Centre for Medium-Range Weather Forecasts (ECMWF), and University of California/US National Center for Atmospheric Research, Climate Extremes lays out the complex dynamics of planetary boundaries and tipping points, providing viewers with a deeper understanding of the powerful forces presently shaping Earth.
Share widely if you agree.
🌿🤝🌍COP16 update
The UN COP16 conference on biodiversity has kept a low profile in most mainstream media this week (I wonder why…?)
In order to reach the commitment from the last COP15 in 2022 to protect 30% of the ocean, a new study finds the world needs to establish 300 large offshore marine protected areas (MPAs) and 190,000 small coastal MPAs by 2030.
Currently, only 8.2% of the ocean is under protection, with just 3% highly protected.
To meet the target, 85 MPAs would need to be created daily starting from next year.
BREAKING from Reuters: UN COP16 nature talks gridlocked as conservation funding trickles in.
🌍Exodus Equator
The world’s hotter climes are getting dryer and less livable.
Much of South America is currently experiencing drought conditions:
A new report co-authored by environmentalist Jonathan Porritt: Exodus Equator aims to:
“raise awareness of the global humanitarian and ecological crises that will be caused by the forced displacement of up to 1 billion people by 2050 – as a direct consequence of accelerating climate change.“
Drought and aridity already influence internal migration worldwide - a new paper in Nature analyses internal migration flows in 72 countries:
“We find that increased drought and aridity have a significant impact on internal migration, particularly in the hyper-arid and arid areas of Southern Europe, South Asia, Africa and the Middle East and South America.“
Air quality in the Indian capital Delhi (population nearly 34 million) reached severe levels this week, with pollution exceeding WHO safe limits by 25-30 times in some areas. Farmers burning crop stubble and Diwali fireworks are being blamed…
“India has the world's highest death rate from chronic respiratory diseases and asthma, according to the WHO. In Delhi, poor quality air irreversibly damages the lungs of 2.2 million or 50 percent of all children.“ — Wikipedia
🇰🇷👥📉Korean population decline
Unless new multi-decade life-extension technologies become rapidly available or immigration policies change, South Korea’s population will experience a radical decrease over the next 50 years:
“Life expectancy of 80, 200K births a year If it stabilizes TODAY, they fell to 16 million Koreans by the end of the century Down from 53 million today“
Related: South Korea is grappling with a growing crisis of "lonely deaths" and social isolation, prompting the government to announce a US$327 million, 5-year plan to combat loneliness.
🧘Memetic savasana
Finally…. the Internet is alive with culture and memes this week… here’s my latest catch:
⌨️Gboard
The latest spoof invention from Google Japan (in what may be the longest running joke in tech for over 10 years): A mobius strip keyboard:
(2 years ago: the world’s longest keyboard)
😅Tempting fate
Anthropic showing some swagger against OpenAI…
🫣Haiku < Sonnet < Opus
3 levels to ASI, illustrated:
…or reinterpreted more darkly:
😈Makin’ Trouble
GenAI video memes are the genesis of a new artform…
The latest creation from Dor Brothers:
🧑🍳Trouble in the kitchen
Meanwhile Gordon Ramsay vs. AI is another zeitgeist AI meme trope: here is the latest episode of Chaos Chef from Lauri Koppela (🎩 @InaBeek for sharing).
🎵Björk on souls and music
Timely clip from everyone’s favourite Icelandic musical genius a few decades back; as relevant today for AI-created music as it was then.
“You can't blame the computer; If there's no soul in the music, it's because nobody put it there"
(video via @Dalos on X)
📚Neuromancer 40th anniversary edition
And finally… one of the greatest works of science fiction, William Gibson’s Neuromancer, is approaching its 40th anniversary. If you like your LLM training tokens printed on dead trees, you can pre-order a limited edition hardback edition (or even a signed “Solander Box" edition…) — would certainly look good on the bookshelf:
🙏🙏🙏 Thanks for making it to the end! Please reach out and get in touch with links and feedback, it’s always appreciated.
More again next week!
Namaste
Ben