{"id":"W4392193191","doi":"10.1186/s13000-024-01464-7","title":"Challenges and barriers of using large language models (LLM) such as ChatGPT for diagnostic medicine with a focus on digital pathology – a recent scoping review","year":2024,"lang":"en","type":"article","venue":"Diagnostic Pathology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":257,"is_retracted":false,"has_abstract":true,"ca_institutions":"SKiN Health","funders":"","keywords":"Interpretability; Context (archaeology); Medicine; Health care; Autonomy; Political science; Artificial intelligence; Computer science","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":[0.1009407,0.001566517,0.00375411,0.01165517,0.002159371,0.01043166,0.004511698,0.00480936,0.004586979],"category_scores_gemma":[0.2998742,0.001573221,0.006530216,0.01073227,0.004168771,0.01329618,0.007707268,0.004382378,0.001104895],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.008015777,"about_ca_system_score_gemma":0.04118826,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01546073,"about_ca_topic_score_gemma":0.02598473,"domain_scores_codex":[0.9162441,0.04696212,0.01937032,0.003095689,0.01297152,0.001356203],"domain_scores_gemma":[0.5195689,0.4347191,0.01569256,0.005340273,0.02354308,0.001136146],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"systematic_review","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0001569647,0.00009666394,0.002404951,0.5454295,0.001379602,0.0006022289,0.01583386,0.0007788451,0.0003694502,0.01228226,0.01292166,0.407744],"study_design_scores_gemma":[0.00003958359,0.000085375,0.001436132,0.8853214,0.002632883,0.0005768886,0.009043536,0.0004255994,0.0003365112,0.004933277,0.09509175,0.00007702939],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"review","genre_gemma":"review","genre_scores_codex":[0.00408432,0.9659591,0.004827097,0.01732329,0.001009746,0.0008893401,0.0004079339,0.00006246306,0.005436717],"genre_scores_gemma":[0.04845402,0.9213195,0.01557893,0.009205294,0.0005364947,0.003499953,0.0006452916,0.00009091331,0.000669518],"genre_candidate":"review","genre_consensus":"review","teacher_disagreement_score":0.1009407,"threshold_uncertainty_score":0.5338318,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1529853727439693,"score_gpt":0.4414231193446899,"score_spread":0.2884377466007206,"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."}}