{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01472599,0.001316426,0.0007466698,0.001317246,0.000451018,0.001477877,0.001429065,0.001417345,0.001740175],"category_scores_gemma":[0.03122258,0.0005083971,0.001297768,0.0005774931,0.0009781525,0.002045208,0.001953169,0.00210506,0.0006130764],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00210249,"about_ca_system_score_gemma":0.00147732,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01477935,"about_ca_topic_score_gemma":0.00995999,"domain_scores_codex":[0.9973954,0.001600595,0.0001402256,0.0004887397,0.0002209954,0.00015406],"domain_scores_gemma":[0.9630289,0.0327988,0.0008409949,0.001621567,0.001187462,0.0005221536],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001758831,0.0003947075,0.09100807,0.0003441016,0.0008156067,0.0001975429,0.001078124,0.7508962,0.004965341,0.00219346,0.003179894,0.1431681],"study_design_scores_gemma":[0.0000349281,0.0002395422,0.006260335,0.00002699328,0.00007076729,0.00004812412,0.00007774276,0.9888905,0.001724815,0.002295761,0.0003062903,0.00002417678],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9148768,0.0007902024,0.07661824,0.001258255,0.00007590643,0.0002160633,0.0009343772,0.001897624,0.003332513],"genre_scores_gemma":[0.9820552,0.0001128556,0.01568826,0.000158612,0.00001982318,0.0001094682,0.001224861,0.00007965646,0.000551279],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01477935,"threshold_uncertainty_score":0.07787943,"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."}}