An Ai2 research scientist says AGI may never emerge because such a concept ignores the physical realities and limits of computation, such as energy constraints
If you are reading this, you probably have strong opinions about AGI, superintelligence, and the future of AI. X: @scaling01 , @sriramk , @tim_dettmers , and @tim_dettmers LinkedIn: Ryan Iyengar and Ali Minai Bluesky: @alexcampolo and @sungkim . Forums: Hacker News X: @scaling01 : I think the ultimate test for AGI is whether AI can debate right now it's fucking terrible at it it keeps moving goalposts and a simple “are you sure” makes it switch positions Sriram Krishnan / @sriramk : Fascinating read on AGI and why we are fundamentally constrained Tim Dettmers / @tim_dettmers : Many people think AI will continue improve towards AGI. In my new blog post, I argue that we will not reach AGI due to physical reasons. Key items discussed: The physical reality of computation Why GPUs will no longer improve Why superintelligence is a fantasy Tim Dettmers / @tim_dettmers : My new blog post discusses the physical reality of computation and why this means we will not see AGI or any meaningful superintelligence: https://timdettmers.com/... LinkedIn: Ryan Iyengar : A solid argument from first principles why the classical definition of AGI won't happen. — Recent models are certainly wildly impressive and useful! … Ali Minai : A very interesting contribution to the growing genre of AGI skepticism. The article makes many important points - especially the physical nature … Bluesky: Alex Campolo / @alexcampolo : “This amplification of bad ideas and thinking exhuded by the rationalist and EA movements, is a big problem in shaping a beneficial future for everyone.” timdettmers.com/2025/12/10/w... Sung Kim / @sungkim : Why AGI Will Not Happen by Tim Dettmers — This blog post is for those who want to think more carefully about these claims and examine them from a perspective that is often missing in the current discourse: the physical reality of computation. — timdettmers.com/2025/12/10/w... Forums: Hacker News : Why AGI Will Not Happen
Context & Ripple Effects
Dettmers’ argument joins a widening disagreement over whether today’s model trajectory should be understood as a path to general intelligence or as a system with persistent limits. Recent coverage has pointed to labs behaving as though models may remain weak at generalization and workplace learning, while an earlier view cast advanced LLMs as early examples of AGI.
The physical-computation framing adds a different objection to timeline debates such as the view that AGI remains a decade away: it challenges whether the classical goal is attainable under real-world energy constraints at all.
First-order effects
- The immediate effect is to shift the AGI discussion from model behavior alone toward the energy and physical resources required to run increasingly capable systems.
- Dettmers’ position gives AGI skeptics a first-principles critique of claims that continued scaling necessarily culminates in a general-purpose intelligence.
Second-order effects
- The argument raises the bar for AI-lab roadmaps: capability claims become harder to separate from the cost, power, and deployment constraints behind them.
- It also reinforces attention to limits on generalization and on-the-job learning, alongside the infrastructure required to sustain further scaling.
Third-order effects
- If this constraint-centered framing gains traction, AI progress may be judged less by an AGI threshold and more by economically deployable capability at bounded energy and compute costs.
- That would favor an industry structure in which infrastructure access and compute efficiency remain central differentiators, rather than treating more scale as a sufficient path to generality.
The trend: The AGI debate is increasingly shifting from abstract capability forecasts toward the physical and economic feasibility of producing and operating those capabilities at scale.