{"id":"W4382318018","doi":"10.1609/aaai.v37i12.26682","title":"Bugs in the Data: How ImageNet Misrepresents Biodiversity","year":2023,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Species Distribution and Climate Change","field":"Environmental Science","cited_by":29,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"Goddard Space Flight Center; Canadian Institute for Advanced Research","keywords":"Benchmarking; Computer science; Artificial intelligence; Set (abstract data type); Machine learning; Image (mathematics); Biodiversity; Wildlife; Object (grammar); Pattern recognition (psychology); Ecology; Biology","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":[],"consensus_categories":[],"category_scores_codex":[0.0115612,0.0007578442,0.0006486925,0.004119093,0.001065021,0.003003145,0.001990681,0.001496485,0.001313429],"category_scores_gemma":[0.06010477,0.0004995441,0.0004351516,0.004494898,0.002265539,0.004084342,0.002052023,0.001295821,0.0009007552],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002187663,"about_ca_system_score_gemma":0.0009520053,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01675626,"about_ca_topic_score_gemma":0.02437962,"domain_scores_codex":[0.9904659,0.003370341,0.0007471765,0.001880536,0.003109055,0.0004270416],"domain_scores_gemma":[0.9672467,0.01829285,0.003724365,0.005482796,0.00483027,0.0004231331],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"observational","study_design_scores_codex":[0.0008635608,0.000260544,0.4602767,0.00117812,0.0005386524,0.001778433,0.006567115,0.03824507,0.006962428,0.02387561,0.1753023,0.2841515],"study_design_scores_gemma":[0.0001040729,0.0002654832,0.3241957,0.001401855,0.0003317671,0.00635613,0.0102089,0.3161183,0.02770496,0.08307217,0.2299617,0.0002788553],"study_design_candidate":"observational","study_design_consensus":"observational","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8130574,0.006324968,0.1030391,0.0163853,0.001954712,0.0004204935,0.02677554,0.006429458,0.02561327],"genre_scores_gemma":[0.9296278,0.0006844673,0.03790376,0.003177668,0.0002538589,0.0001683579,0.02387201,0.001373097,0.002938966],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01675626,"threshold_uncertainty_score":0.06114221,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.2291688641169977,"score_gpt":0.3201386821196012,"score_spread":0.09096981800260354,"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."}}