Oct 2026· Minds and Machines· Vol 36· 0 citations· 88 references
Cybernetics and Technology in Society
Abstract
A persistent ambition in artificial intelligence research holds that large language models (LLMs) will eventually become embodied—through attachment to robots, augmentation with multimodal sensors, integration into vision–language–action architectures, or sheer scale until grounding emerges from training data. This ambition rests on a misunderstanding of embodiment. Embodiment is a constitutive condition of being a self-producing, self-maintaining living system whose continued existence is genuinely at stake in its couplings with the environment. Adding sensors, actuators, or physical instantiation to a computational substrate yields engineered coupling, which is something else entirely. We identify five conditions that any genuinely embodied system must satisfy—autopoietic closure, precariousness, constructive self-determination, intrinsic temporality, and mereological closure—and show that no computational artifact can satisfy any of them, let alone all five. The barrier is categorical incompatibility: the same property that makes a system computational (heteronomous design and maintenance through external specification) is the property that precludes it from being embodied. Coupling artifacts to environments through transducers achieves engineered interaction. Having a body and being a body are different conditions. We close by examining why the embodiment-equivalence claim persists despite its categorical impossibility, and by arguing that the confusion sustains two harms—overselling AI systems for tasks they cannot perform, and misdirecting public concern toward speculative futures while the actual harms of deployed systems accumulate without being addressed. Clarifying the categorical distinction is a precondition for taking the genuine stakes of AI deployment seriously rather than as a philosophical aside.
It is found that Llama 3 delivers comparable quality to leading language models such as GPT-4 on a plethora of tasks, and performs competitively with the state-of-the-art on image, video, and speech recognition tasks.
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.
The results are packaged in the Greenfield Startup Model (GSM), which explains the priority of startups to release the product as quickly as possible, and the need to shorten time-to-market, by speeding up the development through low-precision engineering activities.
Carmine Giardino, Nicolò Paternoster, M. Unterkalmsteiner et al.· IEEE Transactions on Softwar...· 178 citations· ⚡14