{"id":"W4315646142","doi":"10.1148/ryai.220056","title":"Hurdles to Artificial Intelligence Deployment: Noise in Schemas and “Gold” Labels","year":2023,"lang":"en","type":"article","venue":"Radiology Artificial Intelligence","topic":"Big Data and Business Intelligence","field":"Business, Management and Accounting","cited_by":24,"is_retracted":false,"has_abstract":true,"ca_institutions":"Trillium Health Centre; University of Toronto","funders":"Government of Canada; TD Bank","keywords":"Medicine; Software deployment; Noise (video); Artificial intelligence; Software engineering","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.1361471,0.001023184,0.001334609,0.003772846,0.003292737,0.008630696,0.004401157,0.002360886,0.001787953],"category_scores_gemma":[0.3602943,0.001204698,0.001201508,0.005416178,0.005358394,0.006319046,0.007360559,0.003760934,0.00152356],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.004843794,"about_ca_system_score_gemma":0.005120565,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008078015,"about_ca_topic_score_gemma":0.008368664,"domain_scores_codex":[0.826816,0.1075738,0.01468644,0.0179565,0.03103016,0.00193713],"domain_scores_gemma":[0.5770752,0.2830991,0.01698312,0.08231214,0.03829936,0.002231111],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.002754075,0.0008784548,0.2496188,0.005713735,0.001310651,0.001404656,0.0254647,0.02885987,0.02274347,0.07357296,0.1132399,0.4744387],"study_design_scores_gemma":[0.0004785323,0.0006820717,0.1707315,0.003762301,0.0006198494,0.003361656,0.0152589,0.1149059,0.0503417,0.2467523,0.3924903,0.0006150151],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3967988,0.005378801,0.5144411,0.0262343,0.002497132,0.002257361,0.01795865,0.007544614,0.0268892],"genre_scores_gemma":[0.6162313,0.0007053348,0.3377387,0.008659028,0.0004382945,0.003135436,0.02798535,0.002335691,0.00277089],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8638529,"threshold_uncertainty_score":0.7200232,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1293718940637283,"score_gpt":0.3324809187442301,"score_spread":0.2031090246805019,"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."}}