{"id":"W4287889735","doi":"10.18653/v1/2022.findings-naacl.151","title":"Great Power, Great Responsibility: Recommendations for Reducing Energy for Training Language Models","year":2022,"lang":"en","type":"article","venue":"Findings of the Association for Computational Linguistics: NAACL 2022","topic":"Topic Modeling","field":"Computer Science","cited_by":60,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Massachusetts Green High Performance Computing Center; Leukemia and Lymphoma Society of Canada; National Science Foundation","keywords":"Computer science; Energy consumption; Inference; Pace; Cloud computing; Transformer; Efficient energy use; Machine learning; Artificial intelligence; Risk analysis (engineering); Engineering","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002083196,0.0001655406,0.0002886901,0.000221785,0.001075934,0.00009662376,0.0008813959,0.00007044457,0.00001761797],"category_scores_gemma":[0.006130092,0.0001776671,0.0003372334,0.0003963014,0.00001803585,0.0001146006,0.0003695771,0.0001581973,3.511227e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008206541,"about_ca_system_score_gemma":0.0003719381,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004399076,"about_ca_topic_score_gemma":0.000006515796,"domain_scores_codex":[0.997807,0.0001420879,0.0006344541,0.0004975151,0.0005720644,0.0003468265],"domain_scores_gemma":[0.9945863,0.003386992,0.0006935187,0.0003526656,0.0009290152,0.00005149408],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00007566023,0.00005740068,0.0001478751,0.0000318692,0.0001079101,1.840221e-7,0.003689208,0.3642886,0.00008077651,0.6192699,0.01021708,0.002033493],"study_design_scores_gemma":[0.0008835406,0.0001265266,0.00006906575,0.00002022234,0.00004153758,0.000001738814,0.0002668192,0.7034373,0.0001863669,0.2580701,0.03671907,0.0001777189],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.004352561,0.0000808332,0.9853062,0.003563409,0.003013422,0.0009707152,0.001997443,0.0001118973,0.000603544],"genre_scores_gemma":[0.7491319,0.000001394135,0.2455907,0.0005633378,0.0003451702,0.0004908889,0.0005457811,0.00003837499,0.003292366],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.7447793,"threshold_uncertainty_score":0.8275323,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03769841022498763,"score_gpt":0.2914918612409227,"score_spread":0.2537934510159351,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}