GPT-5 and the end of naive AI futurism ⊗ All the world’s polygons ⊗ The future will be boring

GPT-5 and the end of naive AI futurism

According to Dave Karpf, and I tend to agree, “OpenAI’s release GPT-5 feels like the end of something.” It might signal a slowdown from the rapid, exponential progress that AI boosters like Sam Altman have long promised, challenging the optimistic narrative that scaling alone will lead to revolutionary breakthroughs. Despite improvements, current models still cannot replace professionals such as doctors or lawyers, and if future upgrades remain modest, the financial viability of the AI industry is at risk.

Rather than an infinite acceleration going up and to the right, as these boosters have been raving about, the growth of AI technologies seems to follow a typical diffusion-of-innovation pattern, a S-curve, meaning we are witnessing gradual adoption rather than a radical transformation. Investors’ continued support relies heavily on the belief in dramatic future breakthroughs, but with GPT-5’s incremental advances, that story is becoming harder to sustain.

Karpf doesn’t mention this, but compare the most recent GPT-5 to OpenAI’s own OSS 120b, an open weight model available for download. It’s purely speculative for me to say, but looking at the flattening curve of performance and the smaller models, it’s not too far-fetched to think that in a few years you’ll be able to have a GPT-5 equivalent running on your Mac (if said Mac has 64Gb or RAM) while GPT-7 keeps growing in UI, polish, and detailed “thinking,” but not as much in performance.

Then again, the battle has been switching to inference (the “thinking” part), so even if the gap in “intelligence” looks like it’s getting smaller, what will be the gap in “applied intelligence”? (I.e. the IQ is growing slower but the frontier models providers are much better at getting things done than the local options.)

Alternative future; what if, instead of this battle to the death built on billion upon billion of VC money, aiming to ravage field after field, job destruction after job destruction, they’d gone slower? Seems like slowing down to a third or half the breakneck speed they’ve been going could have freaked people out less, permitted more innovation in data quality, respect of rights, energy use, etc. And who knows, maybe they would have had time to figure out a business model?

More → If, to Karpf’s schadenfreudesque near glee at this turn of events, you want more pessimistic but strong takes, you can also read Charlie Warzel’s AI Is a Mass-Delusion Event or Max Read’s A.I. as normal technology (derogatory). I just felt like featuring the lighter-toned one this week.

To hear Altman tell it, the present is only ever prologue. We are living through times of exponential change. Scale is all we need. With enough chips, enough power, and enough data, we will soon have machines that can solve all of physics and usher in a radically different future. […]

Actually-existing-AI today cannot replace your doctor, your lawyer, your professor, or your accountant. If tomorrow’s AI is just a modest upgrade on the fancy chatbot that writes stuff for you, then the financial prospects of the whole enterprise collapse. […]

Look, I am not a business genius. But as I understand it, the point of a business plan is to eventually make more money than you are spending. If your company is breakeven on merchandise, but losses billions per year on the rest of the product line, then that’s a company that is supposed to, y’know, fail.

All the world’s polygons

This week was my yearly pilgrimage public transport commuting to MUTEK Forum. One of the highlights of the event (for me, but judging by the applause I’m not the only one) was Alice Bucknell’s All the world’s polygons, a “lecture-performance exploring ecological simulation in gaming and how predictive technologies can both imagine and constrain future possibilities.” They delivered the lecture while showing Unreal Engine on screen, opening “world files” and going on walkabouts in these worlds. Very impressive.

Sometimes Bucknell’s point of view would stop, swivel up to a screen floating in the world, mimicking a slide from a deck, then the walk (or the flying, the zooming in and out, the scale change) would continue, powerfully illustrating the themes of the text. Going from the map is not the territory, to what gets measured, to digital twins, climate change, more than human lives, NVIDIA’s Earth-2, photogrammetry, asset marketplaces, the Eames’ Power of ten, and lots more.

The polygons link above leads to an essay version of the talk and the quotes below were highlighted there. You can also learn more about the talk here, see a recorded version of the lecture (doesn’t do it justice but worth a look), or read this interview by my friends at HOLO, Playing with fire: Alice Bucknell’s game worlds probe the polycrisis.

As an increasingly integral tool in the industry’s production pipelines, the game engine can be thought of as one huge stage set. Things are sculpted, painted, meticulously lit, and precisely filmed. There is a constant negotiation between environmental complexity and render efficiency that determines how virtual worlds are made; this tension is crucial to how such engines have evolved over the years. […]

As dazzling as the concept of a planet-sized inventory of hi-fidelity natural assets may be, there is an irony to this digital doubling. Recalling the museological impulse to immortalize in glossy vitrines all the world’s life—a desire fueled by colonialism, the invention of taxonomy in the 18th century, and the conceptualization of extinction in the 17th—this ecological doubling feels, inescapably, like a preemptive memento mori to a world in crisis. […]

Here’s the paradox of climate modeling: more accuracy begets more believability, but more specificity begets inaccuracy. In evaluating the accuracy of their future predictions, climate models are assessed on how well they simulate the past. But we are moving into a climate future that is behaving very differently from what we’ve seen before. Earth’s climate is a chaotic pendulum, with multiple systems acting upon each other simultaneously.

§ The future will be boring. “If we don’t start thinking about the future as an extension of the present — and if we don’t develop a dogged focus on the implications of change on the mundane rhythms of everyday life — the future will continue to feel distant, intangible and somehow “other,” and this weakness may grow into a critical failure of our generation.” That’s an opinion piece by Nick Foster for The New York Times, you can also read a short interview he did with The Observer, Predictions are mostly nonsense.

§ How to not build the Torment Nexus. “As industries mature, they tend to get a little boring. And as industries age, and start seeing their own collapse over the horizon, they tend to get… defensive. Bitter. Conservative. Fucking hostile. … Tech, which has always made progress in astounding leaps and bounds, is just speedrunning the cycle faster than any industry we’ve seen before. It’s gone from good vibes, to a real thing, to unicorns, to let’s build the Torment Nexus in record time.”

Futures, Fictions & Fabulations

• When the future pushes back. “In these simulations, each team must consider the opportunities, challenges and surprise combinations amongst themselves then potentially negotiate or collaborate with other teams to influence what happens next. In each round, the teams vote, then, weighted by their influence, or lack thereof, the collective votes drive, or divert, progress.”
• Top 10 Emerging Technologies 2025. “The World Economic Forum’s Top 10 Emerging Technologies Report lists new technologies poised to have a global impact. This collection, developed in collaboration with Frontiers, a publisher of peer-reviewed, open access scientific journals, presents the selection of technologies for 2025 and illustrates their connections to broader social and environmental issues.”

Algorithms, Automations & Augmentations

• In a first, Google has released data on how much energy an AI prompt uses.“Google has just released a technical report detailing how much energy its Gemini apps use for each query. In total, the median prompt—one that falls in the middle of the range of energy demand—consumes 0.24 watt-hours of electricity. … The company also provided average estimates for the water consumption and carbon emissions associated with a text prompt to Gemini.”
• Meta freezes AI hiring after blockbuster spending spree + Zuckerberg again overhauls Meta’s AI efforts. Did he hit his head falling from that surfboard thing? “Meta Platforms has frozen hiring in its artificial-intelligence division after spending months scooping up 50-plus AI researchers and engineers, according to people familiar with the matter.”

Built, Biosphere & Breakthroughs

• Drones to deliver millions of mosquitoes to prevent bird extinctions in Hawai‘i. “After a team of scientists created “reproductively incompatible” male mosquitoes in the lab, the next challenge was to safely deliver them to where they would most effectively suppress mosquito populations in Hawai’i.”
• Our nature positive future. “With a strong focus on measurable outcomes, the Nature Positive movement tracks progress by evaluating species health, ecosystem integrity, and natural processes, offering a comprehensive framework for addressing the biodiversity crisis.”

Asides

• NASA’s Juno Mission Leaves Stunning Legacy of Science at Jupiter. “It revealed a whole different Jupiter than scientists thought they knew. Oddly geometric continent-size storms, in strange yet stable configurations, dance around its poles. Its heaviest matter seems to linger in its skies, while its abyssal heart is surprisingly light and fuzzy. Its innards don’t resemble the lasagnalike layers found in rocky worlds; they look more like mingling swirls of different kinds of ink.” (Via Kottke.)
• BAD LUCK, HOT ROCKS. “In the more than one hundred years since the establishment [of the Petrified Forest park] in 1906, however, some visitors have still been unable to resist the urge to remove wood from the park. Some of these same visitors eventually return their ill gotten souvenirs by mail, accompanied by ‘conscience letters.’ The content of each letter varies, but writers often include stories of misfortune, attributed directly to their stolen petrified wood.” (Via Linkfest #38.)