Writing
Effective Accelerationism: is AI acceleration driven by fundamental laws?
The debate pits cautious long-termism against effective accelerationism — but both miss the more fundamental point. Acceleration isn’t an ideology you pick; it’s a historical law to manage. And the real urgency is building safe AGI before less scrupulous actors do.
Adapted for the site from a LinkedIn essay first published in April 2025, originally developed with Gemini. I’ve kept the argument as written — the field has moved fast since.
A while back I read a Wired piece contrasting cautious “long-termism” on AI with the “effective accelerationist” (e/acc) approach. It captures a real dynamic — but the analysis felt more focused on politics than on the fundamental laws of technological progress we’ve watched play out for decades.
My perspective starts somewhere else: the Law of Accelerating Returns (LOAR — thanks, Kurzweil). The exponential acceleration of technology, and of AI in particular, isn’t so much an ideological choice as an almost inevitable historical trend, intrinsic to the way information evolves. Once you take that seriously, the whole debate reframes.
Why “halt AI” is the wrong instinct
Concern about risk is understandable. But excessive caution that tips into paralysis — “halt AI” — runs into three problems:
- It runs against the nature of progress. LOAR describes a trend, not a policy you can opt out of.
- It’s unrealistic. In a global race, whoever pauses falls behind, potentially ceding ground to less ethically-minded actors.
- It can be actively harmful. It ignores AI’s immense potential to address enormous problems — disease, poverty. The answer to risk isn’t stopping, but accelerating the technological and ethical countermeasures, which are subject to the same accelerating returns.
What e/acc gets right
The accelerationist movement grasps something real: acceleration is powerful, and technology is a fundamental tool for solving problems — including the ones it creates itself. Underneath it runs an optimism about the transformative potential of a beneficial Singularity. I share more of that optimism than not.
The part everyone overlooks: democratization and urgency
Here is the point I think gets lost. LOAR implies that computing power and algorithmic efficiency keep growing exponentially. Soon, building powerful AI — approaching AGI — won’t be the exclusive domain of big tech and states. It will be within reach of small teams, and of uncontrolled actors.
That scenario raises the risk of AI built without adequate precautions, and it raises it exponentially. Which flips the conclusion: the urgent move is not to slow down, but to proactively guide the development of safe, aligned AGI before the dangerous, uncontrolled versions arrive. It’s a race against time for a controllable, beneficial AI.
In that light, a real hope — and an open question — is whether today’s leaders (OpenAI, Google, Anthropic, and the rest) have the vision, the ethics and the resources to steer toward a genuinely beneficial, aligned AGI and ASI. Alongside them, the open-source ecosystem matters as a force for democratization and transparency — a potentially crucial counterweight, even if keeping pace with the computational and economic firepower of the big players remains a steep challenge.
Headlights, not blind driving
The framing I’d push back on is the one that ties e/acc too tightly to a specific ideology. The LOAR view is more fundamental: it’s about observing and managing exponential trends. That’s not “driving blind.” It’s building ever more powerful headlights — including, eventually, through human-machine integration — to handle the rising complexity.
The real challenge
Technological acceleration, as LOAR describes it, is a global reality and looks unstoppable. Seeing it as a mere “blind race” is reductive. The imminent democratization of advanced AI makes one thing critical: not to stop the acceleration, but to actively guide it.
The real challenge is to develop beneficial, controllable AGI before less scrupulous actors can — leaning on responsible leadership, supporting an open ecosystem, and using the technology itself to build the safeguards while we maximize the exponential upside. The direction is a future of expanded intelligence, with the risks managed proactively rather than wished away.
The open question I keep coming back to: is the real urgency to govern the acceleration rather than halt it — and who is best placed to lead it, the tech giants, the open-source community, or some combination of the two?
← danielebianchini.dev