Is AI progress slowing down? ⊗ Future orientation ⊗ The power of critical sensemaking

Is AI progress slowing down?

Great piece by Arvind Narayanan, Benedikt Ströbl, and Sayash Kapoor. They go into some detail on the uncertainty surrounding AI progress, particularly whether model scaling has reached its limit and if inference scaling is the new path forward. The article is split along five main points: whether scaling is actually dead; deference to insiders; progress in capabilities through inference scaling; whether inference scaling the next frontier; and how product development lags capability increase. Read the whole thing for a much better understanding of model scaling vs inference (there’s also a good talk in the Algos section below), the following two points connect to other angles we’ve discussed before.

In their second point, the authors say that we should stop deferring to insiders, which lines up perfectly with Dave Karpf’s article I shared last week, where he said that, like much of his “intellectual camp,” he tends to “calibrate how much grace I offer to incorrect predictions based on how much power was behind them.” They also argue that AI insiders don’t have much of a technical edge when making predictions, and we can’t forget that business and social factors also have a great impact on such forecasts.

Their fifth point, about how “product development lags capability increase” reminds me of a statement I first heard made by Ethan Mollick and have quoted many times, that “even if AI development stopped now, we’d still have ten years of catching up to everything it can do.” In other words, some metrics for AI (and some ad hoc use) have grown by leaps and bounds, but final products and precise, reliable usage still lags behind and likely will for years.

To reiterate, we don’t know if model scaling has ended or not. But the industry’s sudden about-face has been so brazen that it should leave no doubt that insiders don’t have any kind of crystal ball and are making similar guesses as everyone else, and are further biased by being in a bubble and readily consuming the hype they sell to the world. […]

Improvements in exam performance seem to strongly correlate with the importance of reasoning for answering questions, as opposed to knowledge or creativity: big improvements for math, physics and LSATs, smaller improvements for subjects like biology and econometrics, and negligible improvement for English. […]

The lag in product development is compounded by the fact that AI companies have not paid nearly enough attention to product aspects, believing that the general-purpose nature of AI somehow grants an exemption from the hard problems of software engineering.

Future orientation: Understanding hidden narratives in your organization

Reading this piece by Johannes Kleske I was moving my head in agreement all along and highlighting multiple passages. He explains how organizations often have unconscious narratives about the future that shape their culture and decision-making. Many established companies, facing rapid change, typically view the future as a threat, leading to reactive strategies and cultural anxiety. In contrast, those that adopt an aspirational view of the future can foster innovation and proactive behaviours, aligning their teams with a compelling vision. Everyone has a view of the future, whether they think about it or not, and the future will happen regardless of what you do, but determining which visions lead to choices and strategies, and what you want to orient towards makes a huge difference.

This experience illustrates a crucial insight: every organization has an implicit way of using the future that profoundly shapes its culture, decision-making, and behavior. […]

Most critically, these roles tend to be self-reinforcing. Organizations using fear-based future narratives often develop risk-averse cultures that further reinforce defensive positioning. Conversely, organizations with aspirational future narratives tend to attract optimistic, innovation-minded talent that strengthens their future-positive culture. […]

The question isn’t whether the future plays a powerful role in the life of your organization – it does. Rather, the question is whether that role will be shaped by unconscious patterns or by conscious design. The choice—and the opportunity—lies in addressing this critical dimension of organizational development.

The power of critical sensemaking in shaping futures

Another great piece on futures that I agree with and highlighted quite a bit, this one by Anab Jain and Julia Merican at Superflux. They lay out their approach to critical sensemaking, and argue that rather than simply following the common foresight trends, we should engage with uncertainty and explore the connections between seemingly unrelated ideas to uncover new possibilities for change. Critical sensemaking encourages us to ask better questions, embrace the messiness of reality, and develop the agility to adapt our actions based on past experiences.

Critical sensemaking is not just about identifying trends or following the mainstream narrative; it’s about navigating complexity, exploring the grey areas, and drawing connections between seemingly unconnected ideas. It’s about spotting weak signals and engaging with the friction that exists on the edges—because that’s where the real possibility for change lies. […]

Critical sensemaking isn’t just a skill; it’s a way of engaging with the world that allows us to imagine futures that are rich, complex, and full of potential. […]

Critical sensemaking isn’t about having all the answers, but about asking better questions. It’s about developing the mental agility to connect dots that aren’t obviously related. It’s about sitting with uncertainty and allowing it to reveal new pathways forward. It’s a cartography of the future, a way of mapping what we don’t entirely know and can’t always prepare for. […]

We need to adopt a collective stance of epistemic humility if we’re going to meet the demands of both our turbulent present and emergent futures. […]

Attempting to shape our collective futures is to grapple with the fact that futures are elastic, sticky, and indeterminate.

Without you, Patchwork is nothing

When preparing the newsletter, I almost skipped over this one by Jay Springett, since it’s nominally about Midjourney’s new Patchwork product, which I thought might be too niche for too many readers. Thankfully I changed my mind, since Jay actually spends more time on Frank Herbert and Max Barnard’s book Without Me You’re Nothing (1980), which aimed to “to demystify personal computing for a ‘computing-curious’ audience.” A lot of great, prescient, thoughtful takes and wishes by Herbert in there, sketching out today’s computing and knowledge work.

Springett intersects this with Patchwork, which is “a worldbuilding tool with an infinite, zoomable canvas that lets users organise and connect ideas visually—like Figma, but for worldbuilding. It uses ‘scraps’ as building blocks to represent things like characters, events, places, or props. Each scrap can include text, images, and links to show how everything fits together.”

Taken together, we have a 45-year-old vision for “the author’s computer” and a just-arrived tool for worldbuilding that just might be a realisation of it. For all readers, but especially those perhaps less interested in these topics, the title and conclusion are an important reminder that, still today, the computer, and the Patchwork, are nothing without the creators yielding them. (All the quote below are from Herbert’s book and emphasis Jay’s.)

How about calling it the “illusion machine”?” This ability with creative illusion is at once a danger and one of the most attractive characteristics of computers. […]

The same facility that lets you do this with an imagined solar system lets you create exotic spaceships and turn those ships to any desired angle of view on your screen. You can examine the outside skin of your ship or look into its rooms and corridors. You can move that spaceship from one planet to another. Your computer will supply the logical elapsed time according to the program(s) you have supplied. No more laborious figuring of such flight times. […]

Some of the things Hollywood calls science fiction are really comic books for the screen, akin to some early pulp stories in their primitive assumptions and laughable mistakes in science. But it’s obvious that these crude attempts to translate imaginative images into film are still in their infancy.

§ Portable comfort and commodified rebellion: the rise of frictionless conformity. I’m not sure if it’s Gen Z, as the author says, or Alpha, as I would surmise, but a bunch of good hypotheses, the following quote is but one. Also, make sure to click through and check out the author’s incredible Midjourneyed series. “Clothes aren’t fashion anymore; they’re infrastructure. Gen Z doesn’t pick outfits—they select tools to navigate the modern world. Puff jackets and sneakers aren’t stylish; they’re functional default settings. In a globalized system where aesthetic diversity is crushed by mass production, these items are the universally compatible operating system for human bodies. Practical, unobtrusive and designed to work in any environment, these clothes are less about identity and more about surviving the interface of late-stage capitalism.”

§ An interview with Yatú Espinosa. Just a nice chat with creative folks doing fun stuff with an inspiring creative philosophy of learning. Also, contains a bunch of links to dig into. “There’s something more rewarding about building on a concept rather than moving on from it. It also aligns with a recurring theme in my life. For instance, we call our home Playground, the studio Daycare, and past spaces have been named School and Factory. I even refer to my partner’s place as Campgrounds because of her patio. It’s a way of reshaping language to create a metaphorical framework that adds depth and meaning to spaces.”

§ Science journalism becomes plain old journalism. The link is to one of the many pieces part of the yearly Nieman Journalism Lab Predictions for Journalism. “For too long, science journalism has been treated as something distinct, something extra — the domain of specialists writing for audiences who are already deeply interested in and informed about science. This is bad.”

Holiday long(er) reads

A few promising longer reads I haven’t gotten to and that will likely be lost to the vortex coming back from the break.

• An Age of Hyperabundance | n+1
• The New Artificial Intelligentsia | Los Angeles Review of Books
• Microsoft’s Mustafa Suleyman on what the industry is getting wrong about AGI | The Verge
• The confusing reality of AI friends | The Verge

Futures, Fictions & Fabulations

• deliberate oxymorons: an interview with Bruce Sterling (part 1). “In the first part of this two-part interview, OG cyberpunk Bruce Sterling discusses the pursuit of deliberate oxymorons as a creative strategy, worldbuilding in the context of history and futurity, Berlusconi on the moon, and much, much more.”
• New report: How do we explore the future with strategic foresight? “The report compares Sweden's work on strategic foresight with international forerunners such as Singapore, Finland and Great Britain. The results show that Sweden has a lot to gain from institutionalizing foresight work and integrating it into policy development and innovation strategies. Countries that have established strong foresight functions clearly show how this work can strengthen competitiveness and innovative capacity.”
• From pioneer to futurist: A conversation with Melissa Sterry. “Exemplifying the pioneer-to-futurist archetype, she broke new ground in sustainable design, biomimetics, and systems thinking—long before formally embracing her role as a biofuturist. Melissa’s work also delves into profound questions about societal change and collapse: What happens when systems fail? How do we prepare and adapt?”

Algorithms, Automations & Augmentations

• Genesis looks pretty mind boggling. “A universal physics engine” + “lightweight, ultra-fast, pythonic, and user-friendly robotics simulation platform” + “photo-realistic rendering” + “generative data engine”. However, I’ve seen some rumblings that it’s just aspirational and fake, tbc. I’m sharing anyway, for those who want to track that debate, and because natural language AI as an interface on top of something less hallucinatory, which I believe this purports to be, will be a trend to keep an eye on.
• Google debuts new AI video generator Veo 2 claiming better audience scores than Sora. “A better understanding of real-world physics and the nuances of human movement and expression.” … “Veo 2 also understands the language of cinematography: Ask it for a genre, specify a lens, suggest cinematic effects and Veo 2 will deliver — at resolutions up to 4K.”
• Andrew Ng explores the rise of AI agents and agentic reasoning. He’s not the most dynamic speaker, but a really useful overview of the different concepts behind “agents.”

Built, Biosphere & Breakthroughs

• Trespassing for the Common Good. “In England, a movement is growing to defy enclosure by trespassing on private land. These transgressions aim to model a more democratic relationship with nature.”
• Nimble electric trucks are supercharging African trade. Looove this kind of thing! “In Rwanda, farmers often watch their harvest spoil before it can reach the market. A fleet of simple, efficient trucks is changing that.” … “Launched in Rwanda in 2021, Ox trucks are almost 10 times cheaper to run than existing alternatives.”
• How slowing Amazon’s deforestation improved Brazil’s health. “Policy changes in Brazil that have disincentivized deforestation of the Amazon for agriculture are credited with reducing hospitalizations and wildfire deaths by the thousands.”
• Squid-bone sponge found to soak up 99.8 per cent of microplastics. “Researchers from the University of Wuhan and Guangxi University produced the fibrous foam by combining cellulose fibres from cotton and the tough biopolymer chitin, which forms a squid's skeleton.”

Asides

• I built Apple’s 1980s iPad concept!. If you like watching “crafting stuff” videos (and electronics, computing), this is an excellent one. “This project has been a dream come true! As a designer, I've always been inspired by groundbreaking concepts, and this time, I challenged myself to recreate one of the most iconic and unrealized prototypes: Hartmut Esslinger’s Apple FlatMac.”
• Building Stuff. A reader sent me this series he’s worked on, looks super promising (on NOVA). “Engineering is all around us, and we humans have been doing it forever. But how does it actually work? Find out by watching some of the most creative and innovative folks in the game build stuff that helps extend our range, amplify our abilities, and alter our environment for the better. Experience the ups and downs with engineers as they design, build, and iterate their way through challenges, inspiring the inner ‘maker’ in all of us.”
• Aerial Foodscapes. “For his book Feed the Planet: A Photographic Journey to the World’s Food, George Steinmetz travelled the world with his drone, spending ‘a decade documenting food production in more than 36 countries on 6 continents, 24 US states, and 5 oceans’.”