Oct 2026· Zenodo (CERN European Organization for Nuclear Research)· 2 citations· 2 references
Green IT and Sustainability
Abstract
Abstract. Claims that artificial intelligence approaches a planetary thermodynamic boundary routinely conflate four quantities: the Landauer lower bound per irreversible bit erasure (Q1), measured hardware efficiency η (Q2), physically available electricity (Q3), and renewable-energy-certificate accounting (Q4). This paper separates them, corrects arithmetic and unit errors from an earlier draft, and replaces point predictions with a scenario band. With 2024 renewable generation of 8,500 TWh and an assumed CMOS η = 10⁷, the energy-budget capacity is 1.07×10³³ erased bits (1.07×10⁴⁰ at η = 1); 2030 scenarios give 2.47–3.18×10³³ bits. Current AI electricity (50 TWh in 2024) is 0.59% of same-year renewable generation. The earlier '5–8 years' claim is one point in a band running from ~4 to ~33 years depending on an unestablished growth assumption. RECs are accounting instruments, not physical supply. The paper closes with an explicit list of statements the evidence cannot support and a reporting checklist. Contents. Markdown source, DOCX and PDF; three figures; reproducible calculation script and the corrected section draft. Companion artifacts. This record is part of the SMMB long-term program together with P1 (two-node systems prototype) and P2 (control-theory assumptions, counterexamples and a fair-baseline negative result). Cross-paper claims are limited to each paper's scope; see the claim-boundary table in the manuscript. Status. Working draft; not peer reviewed. Published as a preprint for citation and reproducibility. No venue deadline is implied.
This paper investigates developing real-world perception for active agents, proposes Gibson Environment for this purpose, and showcases a set of perceptual tasks learned therein.
F. Xia, Amir Zamir, Zhi-Yang He et al.· 2018 IEEE/CVF Conference on...· 1,069 citations· ⚡112
GAOKAO-Bench is introduced, an intuitive benchmark that employs questions from the Chinese GAOKAO examination as test samples, including both subjective and objective questions that contribute a robust evaluation benchmark for future large language models and offers valuable insights into the advantages and limitations...
Xiaotian Zhang, Chun-yan Li, Yi Zong et al.· arXiv.org· 216 citations· ⚡17
An algorithm is proposed which leverages disentangled semantic factors to generate adversarial perturbation by altering controlled semantic attributes to fool the learner towards various "adversarial" targets.
Haonan Qiu, Chaowei Xiao, Lei Yang et al.· arXiv.org· 208 citations· ⚡17
This work proposes a HALO that directly maximizes the utility of generations instead of maximizing the log-likelihood of preferences, and suggests that the best loss depends on the inductive biases most appropriate for a given setting, an oft-overlooked consideration.
This work investigates the possibilities of using LLMs in a resume screening setting via a document retrieval framework that simulates job candidate selection and finds that the MTEs are biased, significantly favoring White-associated names in 85% of cases and female-associated names in only 11.1% of cases.
This work shows that orders of magnitude enhancement in performance could be obtained by a combination of hardware improvements and tight quantum-HPC integration and introduces high-performance architectures for quantum-probabilistic computing with custom-designed accelerators to tackle today's industry-scale classical...
Masoud Mohseni, Artur Scherer, K. Johnson et al.· arXiv.org· 121 citations· ⚡9