{"id":"W3184097711","doi":"10.1038/s42256-021-00370-7","title":"Governing AI safety through independent audits","year":2021,"lang":"en","type":"article","venue":"Nature Machine Intelligence","topic":"Ethics and Social Impacts of AI","field":"Social Sciences","cited_by":221,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of British Columbia","funders":"Economic and Social Research Council; Engineering and Physical Sciences Research Council; National Institute of Standards and Technology; National Science Foundation","keywords":"Audit; Business; Risk analysis (engineering); Accounting","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["research_integrity","insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.001159586,0.0001715597,0.0002278661,0.00002870142,0.0009970887,0.0002975462,0.00049863,0.0007366698,0.00135066],"category_scores_gemma":[0.005129583,0.0001674396,0.0001446801,0.0005694683,0.0002554679,0.0005439994,0.0001723767,0.002526085,0.0001791723],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002428668,"about_ca_system_score_gemma":0.0007484938,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003344061,"about_ca_topic_score_gemma":0.01447806,"domain_scores_codex":[0.9975147,0.0003088746,0.0002834492,0.0003810539,0.001034656,0.0004772804],"domain_scores_gemma":[0.9982415,0.0004858064,0.0001218919,0.0002698817,0.0007042455,0.0001767037],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"not_applicable","study_design_scores_codex":[0.00002942862,0.0001160816,0.002234777,0.00002272754,0.0000643906,0.0001533827,0.02768366,0.0001419872,0.000199211,0.9301246,0.00564811,0.0335816],"study_design_scores_gemma":[0.0001063769,0.00003795165,0.001857806,0.00009120035,0.00002605648,0.000008986883,0.005080191,0.0001067599,0.004868495,0.1509444,0.8364486,0.0004231781],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"other","genre_gemma":"empirical","genre_scores_codex":[0.006028078,0.0122927,0.03731886,0.1346819,0.004569529,0.0003653678,0.0001162741,0.0003220013,0.8043053],"genre_scores_gemma":[0.9774554,0.002763978,0.001092326,0.01179761,0.0007642367,0.000002814666,0.00002271403,0.00001946028,0.006081504],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9714273,"threshold_uncertainty_score":0.9997751,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02180139953642077,"score_gpt":0.3918023780780409,"score_spread":0.3700009785416201,"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."}}