We demonstrate a transparent and reproducible system that estimates how current AI technologies can perform job-related tasks and how this translates into occupation-level exposure and replacement, leveraging open Large Language Models (LLMs). The demo focuses on two indices: TEAI (Task Exposure to AI) and TRAI (Task Replacement by AI). Terminator Economy combines (i) a multi-LLM panel with a consensus mechanism to estimate task-level exposure and replacement rates, (ii) an aggregation of tasks by occupation weighted by relevance, importance, and frequency provided by official taxonomies. We present an online web demo that allows users to explore exposure/replacement for each occupation at the task level, and audit model rationales. We release code and data to foster benchmarking of AI progress over time.
Colombo, E., Mercorio, F., Mezzanzanica, M., Serino, A., Seveso, A. (2027). Terminator Economy: Assessing Jobs and Tasks Exposure to AI. In Machine Learning and Knowledge Discovery in Databases. Applied Data Science Track, Demo Track and Industrial Track European Conference, ECML PKDD 2026, Naples, Italy, September 7–11, 2026, Proceedings, Part X (pp.368-371) [10.1007/978-3-032-37685-5_36].
Terminator Economy: Assessing Jobs and Tasks Exposure to AI
Mercorio F.;Mezzanzanica M.;Serino A.;Seveso A.
2027
Abstract
We demonstrate a transparent and reproducible system that estimates how current AI technologies can perform job-related tasks and how this translates into occupation-level exposure and replacement, leveraging open Large Language Models (LLMs). The demo focuses on two indices: TEAI (Task Exposure to AI) and TRAI (Task Replacement by AI). Terminator Economy combines (i) a multi-LLM panel with a consensus mechanism to estimate task-level exposure and replacement rates, (ii) an aggregation of tasks by occupation weighted by relevance, importance, and frequency provided by official taxonomies. We present an online web demo that allows users to explore exposure/replacement for each occupation at the task level, and audit model rationales. We release code and data to foster benchmarking of AI progress over time.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.


