{"id":"W4388287351","doi":"10.1136/bjo-2023-324438","title":"Capabilities of GPT-4 in ophthalmology: an analysis of model entropy and progress towards human-level medical question answering","year":2023,"lang":"en","type":"article","venue":"British Journal of Ophthalmology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":84,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université du Québec à Montréal; Institut universitaire en santé mentale de Montréal; Hôpital Maisonneuve-Rosemont; Université de Montréal; Institut Universitaire en Santé Mentale de Québec; Centre Hospitalier de l’Université de Montréal","funders":"National Institute for Health and Care Research","keywords":"Medicine; Optometry; Ophthalmology; Medical physics; Computational biology","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":[],"consensus_categories":[],"category_scores_codex":[0.001720736,0.00009952766,0.0007357412,0.0007856571,0.00004439333,0.000007875098,0.0001381547,0.0002474079,0.000206919],"category_scores_gemma":[0.0006982821,0.0001087181,0.0001222108,0.0006318623,0.0005249233,0.0001407223,0.00003886656,0.0003258698,4.670213e-7],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00007822606,"about_ca_system_score_gemma":0.0004016386,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003414526,"about_ca_topic_score_gemma":0.00008751424,"domain_scores_codex":[0.9978072,0.0002377803,0.001141132,0.0001892795,0.0003756986,0.0002489209],"domain_scores_gemma":[0.9986262,0.0001494208,0.0003916397,0.0001359056,0.0005104326,0.0001863535],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0003572895,0.0007534696,0.9391578,0.0004646382,0.0003553062,0.006875861,0.003740411,0.002827619,0.00129862,0.0005127998,0.00002402377,0.04363219],"study_design_scores_gemma":[0.0002239619,0.00141183,0.9149147,0.0005350218,0.0002755822,0.04869851,0.003759549,0.02465956,0.001637744,0.003778425,0.000001009733,0.0001040843],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9980987,0.0008811951,0.0001573034,0.0004407596,0.0001710154,0.0001318519,0.00001697718,0.000006530438,0.0000956551],"genre_scores_gemma":[0.9987772,0.0002377152,0.0007910307,0.00001227921,0.00008315576,0.00000842073,0.00002864496,0.00001211248,0.0000494629],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.04352811,"threshold_uncertainty_score":0.5161765,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.19802124344265,"score_gpt":0.4740761825352103,"score_spread":0.2760549390925604,"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."}}