Massed Muddler Intelligence ⊗ Active vs passive learning ⊗ The staggering impacts of the cloud

Massed Muddler Intelligence

Reading this piece by Venkatesh Rao, get ready to spend some time, Google DuckDuckGo some terms and memorise some acronyms, this is quite the ride. To paraphrase the whole thing: Altman (OpenAI) and co; let’s scale AI by growing ours into a God. Rao; let’s scale by growing a society of smaller AIs “combining embodiment, boundary management, temporality, and personhood elements.”

Here by muddling he means something like ‘figure stuff out by trial and error.’ Rao argues that AI can’t scale to the size/intelligence the current AI bigwigs have in mind, it would be like building a brick and mortar skyscraper, it’s impossible. You need a structural equivalent to reinforced concrete. To achieve this, he proposes lots of smaller AIs with specific characteristics, operating under emerging rules more similar to liberal democracies than to the autocratic-like mega-sized LLMs AGIers dream of.

Rao builds his argument over multiple sections, including AI history, biological precedents, and “muddling doctrines.” It’s a fascinating read and he’s already teasing some followups on some of the specifics. I’m not quite sure how plausible that model is, certainly feels like a more likely road than AGI followed by the singularity, but I’m not technically knowledgeable enough to assess that. Regardless, great read, you should put some time aside to read through.

The MMI vision I’m going to outline could be considered “liberal democracy for mixed human-AI agent systems.” Rather than the autocratic idea of “alignment” associated with “AGI,” MMIs will call for something like the emergent mutualist harmony that characterizes functional liberal democracies. You don’t need an “alignment” theory. You need social contract theory. […]

This is a new, oozy kind of intelligence we are building with for the first time. In my opinion, you can’t agentify and scale it with BDI [Belief-Desire-Intention] type architectures. We’re in new regimes, dealing with fundamentally new building materials and aiming for new scales (orders of magnitude larger than anything imagined in the 1990s). […]

The physicality of computation in general, and AI computation in particular, is consequential and must be both respected and leveraged to build powerful AIs that scale well. AIs must inhabit real atoms, live in real time, and deal with the second law of thermodynamics. And these facts have to matter architecturally. […]

When you put a lot of them together using a mix of hard coordination protocols (including virtual-economic ones) and softer cultural protocols, you get a massed muddler intelligence, or MMI. Market economies and liberal democracies are loose, low-bandwidth examples of MMIs that use humans and mostly non-AI computers to scale muddler intelligence. The challenge now is to build far denser, higher bandwidth ones using modern AI agents.

Active vs passive learning

I’ve often written about the importance of naming things and the power of language, this is a low-key example of that. Morgan Housel writes about active and passive learning and the value of the latter. I found it intriguing because he basically used the words in the opposite way to what I expected. Since I read it right after the piece above, I might have called these two forms of learning ‘purposeful’ vs ‘muddling.’

Housel calls school-based learning active and self-thought passive because you “let your mind wander with no intended destination. You read and learn broadly, talk to people from various backgrounds, and stumble haphazardly across topics you had never considered but spark your curiosity, often because it’s the topic you happen to need at that specific time of your life.” Good short read also on the importance of curiosity transdisciplinarity.

David Senra recently summarized Jeff Bezos’s mindset: “If you’re paying attention, the whole world is a classroom.” A classroom of passive learning. […]

Read broadly, watch broadly, discuss broadly, learn broadly. […]

So many of these employees will do better work if they are given time to think, learn, ponder, discuss, and let their minds roam. But they often can’t, because so many bosses expect them to be at their desk, typing, moving a mouse, 40 hours a week until age 65.

The staggering ecological impacts of computation and the cloud

Not a new argument, there have been a number of studies and articles on the, yes, staggering ecological impacts of tech, I liked this one because the author doesn’t stop at just one or two factors and covers energy use, cooling specifically, carbon emissions, water usage, noise pollution, and e-waste. Although that last one is just a quick glance over, water for once was looked at in a bit more detail than just saying “they use billions of gallons of water.”

Reading this made me think that we have the concept of digital twins, the virtual copies of real things, meant to analyse and simulate. A medical digital twin of your body, the digital twin of a building to analyse AC circulation or elevator traffic, etc. We should also talk more about the reverse; ‘externality twins,’ the physical existence and impacts of the virtual stuff we use every day. I also noticed the use of the term “factory-libraries” for data centers, which echoes Jensen Huang in the last issue, who said they were working on “AI factories.”

… a technique called “free cooling.” However, network signal latency issues make this dream of a [high latitude] haven for green data centers largely untenable to meet the computing and data storage demands of the wider world. […]

The electricity utilized by data centers accounts for 0.3 percent of overall carbon emissions, and if we extend our accounting to include networked devices like laptops, smartphones, and tablets, the total shifts to 2 percent of global carbon emissions. […]

Historian Nathan Ensmenger writes that a single desktop computer requires 240 kilograms of fossil fuels, 22 kilograms of chemicals, and 1,500 kilograms of water to manufacture. […]

The Cloud is both cultural and technological. Like any aspect of culture, the Cloud’s trajectory — and its ecological impacts — are not predetermined or unchangeable. Like any aspect of culture, they are mutable.

Related → Kate Crawford in Nature with Generative AI’s environmental costs are soaring — and mostly secret.

§ Two articles I’m putting aside and that might not make it in the newsletter otherwise, very long read by Shannon Mattern with a political ecology of the repair manual (with quite a few illustrations), and at NOEMA, the extraordinary lives of Coast Redwoods.

§ Work is a place. I’m sure quite a few readers are working from home, this personal insight by Tom Critchlow might be a useful line of thought for you. Tom recently shared a malaise and feeling stuck over the last couple of years, he now believes it’s because he was solely working from home, instead of also coworking and working from client offices.

Futures, Fictions & Fabulations

Nobody owns the technofuture
“Through this blog series, we are exploring what the world can look like when technology is designed and deployed for the benefit of all. These broad, near-future speculative pieces are designed to de-center dominant narratives and challenge us all to realize that things can be different. These are not alternative realities, they are possible futures.”

Enhancing general education with strategic foresight
“As currently designed, general education curricula do not provide students opportunities to demonstrate integrative learning related to the multiple disciplines they have studied. They also do not prepare students to systematically explore, think about, and gain insights from the future to make better decisions in the present. Using strategic foresight as the framework for a general education capstone course would address these gaps.”

The Fandom Forecast 2024
“EQL surveyed more than 900 fans to find out why they buy, how they buy, where they interact with the brands they love most, and how much time and money they plan to spend on products they're passionate about in 2024. Download the full report to see all our exclusive insights, plus predictions on the top 4 things brands can do this year to ignite fandom in their communities.”

Algorithms, Automation, Augmentation

The AI project pushing local languages to replace French in Mali’s schools
“RobotsMali uses ChatGPT, Google Translate, and other AI tools in hopes of helping young students learn faster and stay in school. The initiative has gained importance since the country’s military government removed French as the national language in 2023. Students say they enjoy learning in their local languages, but worry it might limit their global opportunities.”

“Forty-one per cent of architects now using AI” says RIBA report
“Additionally, 43 per cent of those architects using AI believe it has improved the efficiency of their design process. Among the ways these architects are currently using the technology is for the automation of admin-based tasks and to reduce the carbon footprint of projects in tandem with digital twins.”

Robots who share your accent are more trusted, study shows
“While robots speaking in standardized accents are widely seen as more intelligent and competent, some researchers argue that they could be perceived as more familiar and comforting if they talk in more regional accents and dialects — possibly making them seem more trustworthy.”

Asides

• ⏳ 🦴 🧬 🇷🇺 On the Trail of the Denisovans. “What the Denisovans lack in fossils they make up for in DNA. Geneticists have been able to extract bits of genetic material from teeth and bones dating back 200,000 years. They have found genetic clues in the dirt of cave floors. And billions of people on Earth carry Denisovan DNA, inherited from interbreeding.”
• 🤬 😨 🥽 🧬 A Study Found Microplastics in Every Single Human Placenta Tested. “The adverse health impacts of nano- and microplastics is not fully known, but research is emerging that shows the damage these plastic chemicals can have on human cells and tissue. The spreading reach of microplastics in nature—from drinking water to our human bodies—makes them unavoidable.”
• 👏🏼 🧲 🥽 🇮🇪 Fionn Ferreira’s Ground Breaking Solution in Microplastic Removal. “[His] solution involves the use of a unique magnetic mixture known as Ferrofluid. This innovative concoction binds to microplastic particles, separating them from water and allowing for their removal using magnets.”
• 👏🏼 🤩 🏆 📚 🇺🇸 How Mychal Threets Became America’s Favorite Librarian. “His compassion radiates through the screen as he promotes positivity, belonging and “library joy.” In a country faced with a dire literacy crisis, coinciding with regressive and discriminatory book bans and attacks on third spaces from conservatives, Threets’ work to make reading more accessible is needed now more than ever.” (Via RADAR.)
• 👏🏼 🌳 🇨🇴 How a Colombian City Cooled Dramatically in Just Three Years. “With ‘green corridors’ that mimic the natural forest, the Colombian city is driving down temperatures — and could become five degrees cooler over the next few decades.”
• 😨 🌊 🇺🇸 Cities Aren’t Prepared for a Crucial Part of Sea-Level Rise: They’re Also Sinking. “Previous projections have studied geocentric sea-level rise, or how much the ocean is coming up along a given coastline. This new research considers relative sea-level rise, which also includes the vertical motion of the land. That’s possible thanks to new data from satellites that can measure elevation changes on very fine scales along coastlines.”
• 🌳 🌾 🤓 Biochar explainer at MIT’s Climate Portal. “In the right conditions, biochar can make crops grow larger and faster. Biochar rarely contains many nutrients itself, but it can change the soil in ways that improve “nutrient availability”: the ability of soil to hold onto nutrients like nitrogen, phosphorus, and calcium in forms that plants can easily absorb. It can also help restore soils that have become too acidic, or contaminated with certain heavy metals.”