Add new classifiers 20241216

company_linkage/sql/ai_company_papers.sql

@@ -13,7 +13,9 @@ SELECT DISTINCT
   year,
   cv,
   nlp,
-  robotics
+  robotics,
+  ai_safety,
+  llm
 FROM
   unnested_rors
 INNER JOIN
@@ -26,7 +28,9 @@ SELECT
   year,
   cv,
   nlp,
-  robotics
+  robotics,
+  ai_safety,
+  llm
 FROM
   staging_ai_companies_visualization.org_name_matches
 INNER JOIN

company_linkage/sql/ai_publications.sql

@@ -4,9 +4,17 @@ WITH
     merged_id,
     cv_filtered,
     nlp_filtered,
-    robotics_filtered
+    robotics_filtered,
+    ai_safety_predictions.preds = 1 AS ai_safety,
+    COALESCE(llm_classifier_predictions.label, FALSE) AS llm
   FROM
-    gcp-cset-projects.article_classification.predictions
+    article_classification.predictions
+  LEFT JOIN
+    almanac_classifiers.llm_classifier_predictions
+  USING (merged_id)
+  LEFT JOIN
+    ai_safety_datasets.ai_safety_predictions
+  USING (merged_id)
   WHERE
     ai_filtered = TRUE OR cv_filtered = TRUE OR nlp_filtered = TRUE OR robotics_filtered = TRUE),
   ror AS (
@@ -29,7 +37,9 @@ WITH
     org_name,
     cv_filtered as cv,
     nlp_filtered as nlp,
-    robotics_filtered as robotics
+    robotics_filtered as robotics,
+    ai_safety,
+    llm
   FROM
     literature.affiliations
     -- if they're AI papers

company_linkage/sql/check_paper_counts_exist.sql

@@ -7,5 +7,7 @@ WHERE
   OR robotics_pubs IS NULL
   OR cv_pubs IS NULL
   OR nlp_pubs IS NULL
+  OR ai_safety_pubs IS NULL
+  OR llm_pubs IS NULL
   OR ai_pubs_in_top_conferences IS NULL
-  OR all_pubs IS NULL
+  OR all_pubs IS NULL

company_linkage/sql/initial_paper_visualization_data.sql

@@ -6,6 +6,8 @@ WITH
     cv,
     nlp,
     robotics,
+    ai_safety,
+    llm,
     ref_id
   FROM
     staging_ai_companies_visualization.ai_company_papers
@@ -20,6 +22,8 @@ WITH
     cv,
     nlp,
     robotics,
+    ai_safety,
+    llm,
     ref_id,
     year
   FROM
@@ -37,7 +41,9 @@ WITH
     COUNT(DISTINCT merged_id) AS ai_citation_count,
     COUNT(DISTINCT IF(cv, merged_id, null)) AS cv_citation_count,
     COUNT(DISTINCT IF(nlp, merged_id, null)) AS nlp_citation_count,
-    COUNT(DISTINCT IF(robotics, merged_id, null)) AS robotics_citation_count
+    COUNT(DISTINCT IF(robotics, merged_id, null)) AS robotics_citation_count,
+    COUNT(DISTINCT IF(ai_safety, merged_id, null)) AS ai_safety_citation_count,
+    COUNT(DISTINCT IF(llm, merged_id, null)) AS llm_citation_count
   FROM
     add_year
   GROUP BY
@@ -61,7 +67,15 @@ all_cited as
   ARRAY_AGG(STRUCT(year,
       robotics_citation_count as num_papers)
   ORDER BY
-    year) AS robotics_citation_count_by_year
+    year) AS robotics_citation_count_by_year,
+  ARRAY_AGG(STRUCT(year,
+      ai_safety_citation_count as num_papers)
+  ORDER BY
+    year) AS ai_safety_citation_count_by_year,
+  ARRAY_AGG(STRUCT(year,
+      llm_citation_count as num_papers)
+  ORDER BY
+    year) AS llm_citation_count_by_year
 FROM
   by_year
 GROUP BY
@@ -71,7 +85,9 @@ SELECT
   ai_citation_count_by_year,
   cv_citation_count_by_year,
   nlp_citation_count_by_year,
-  robotics_citation_count_by_year
+  robotics_citation_count_by_year,
+  ai_safety_citation_count_by_year,
+  llm_citation_count_by_year
 FROM
   high_resolution_entities.aggregated_organizations
 LEFT JOIN

company_linkage/sql/initial_visualization_data.sql

@@ -8,7 +8,9 @@ WITH
     merged_id,
     cv,
     nlp,
-    robotics
+    robotics,
+    ai_safety,
+    llm
   FROM
     staging_ai_companies_visualization.ai_company_papers),
   rortable AS (
@@ -18,8 +20,9 @@ WITH
     COUNT(DISTINCT merged_id) AS ai_pubs,
     COUNT(DISTINCT CASE WHEN cv IS TRUE THEN merged_id END) as cv_pubs,
     COUNT(DISTINCT CASE WHEN nlp IS TRUE THEN merged_id END) as nlp_pubs,
-    COUNT(DISTINCT CASE WHEN robotics IS TRUE THEN merged_id END) as robotics_pubs
-
+    COUNT(DISTINCT CASE WHEN robotics IS TRUE THEN merged_id END) as robotics_pubs,
+    COUNT(DISTINCT CASE WHEN ai_safety IS TRUE THEN merged_id END) as ai_safety_pubs,
+    COUNT(DISTINCT CASE WHEN llm IS TRUE THEN merged_id END) as llm_pubs
   FROM aipubs
   GROUP BY
     CSET_id)
@@ -51,7 +54,9 @@ SELECT
   COALESCE(ai_pubs, 0) as ai_pubs,
   COALESCE(cv_pubs, 0) as cv_pubs,
   COALESCE(nlp_pubs, 0) as nlp_pubs,
-  COALESCE(robotics_pubs, 0) as robotics_pubs
+  COALESCE(robotics_pubs, 0) as robotics_pubs,
+  COALESCE(ai_safety_pubs, 0) as ai_safety_pubs,
+  COALESCE(llm_pubs, 0) as llm_pubs
 FROM
   high_resolution_entities.aggregated_organizations
 LEFT JOIN

company_linkage/sql/visualization_data_with_by_year.sql

@@ -9,7 +9,9 @@ WITH
     year,
     cv,
     nlp,
-    robotics
+    robotics,
+    ai_safety,
+    llm
   FROM
     staging_ai_companies_visualization.ai_company_papers),
   rortable AS (
@@ -20,7 +22,9 @@ WITH
     COUNT(DISTINCT merged_id) AS ai_pubs,
     COUNT(DISTINCT CASE WHEN cv IS TRUE THEN merged_id END) as cv_pubs,
     COUNT(DISTINCT CASE WHEN nlp IS TRUE THEN merged_id END) as nlp_pubs,
-    COUNT(DISTINCT CASE WHEN robotics IS TRUE THEN merged_id END) as robotics_pubs
+    COUNT(DISTINCT CASE WHEN robotics IS TRUE THEN merged_id END) as robotics_pubs,
+    COUNT(DISTINCT CASE WHEN ai_safety IS TRUE THEN merged_id END) as ai_safety_pubs,
+    COUNT(DISTINCT CASE WHEN llm IS TRUE THEN merged_id END) as llm_pubs
   FROM  aipubs
    WHERE year IS NOT NULL
   GROUP BY
@@ -46,6 +50,14 @@ WITH
         robotics_pubs as num_papers)
     ORDER BY
       year) AS robotics_pubs_by_year,
+   ARRAY_AGG(STRUCT(year,
+        ai_safety_pubs as num_papers)
+    ORDER BY
+      year) AS ai_safety_pubs_by_year,
+   ARRAY_AGG(STRUCT(year,
+        llm_pubs as num_papers)
+    ORDER BY
+      year) AS llm_pubs_by_year
   FROM
     high_resolution_entities.aggregated_organizations
   LEFT JOIN
@@ -59,7 +71,9 @@ SELECT
   ai_pubs_by_year,
   cv_pubs_by_year,
   nlp_pubs_by_year,
-  robotics_pubs_by_year
+  robotics_pubs_by_year,
+  ai_safety_pubs_by_year,
+  llm_pubs_by_year
 FROM
   staging_ai_companies_visualization.initial_visualization_data
 LEFT JOIN

web/gui-v2/src/components/DetailViewPublications.jsx

@@ -126,6 +126,8 @@ const DetailViewPublications = ({
     { text: "Computer vision", val: "cv_publications" },
     { text: "Natural language processing", val: "nlp_publications" },
     { text: "Robotics", val: "robotics_publications" },
+    { text: "Large language models", val: "llm_publications" },
+    { text: "AI safety", val: "ai_safety_publications" }
   ];
 
   return (

web/gui-v2/src/static_data/table_columns.js

@@ -261,6 +261,20 @@ const columnDefinitions = [
     category: "publications",
     ...generateSliderColDef("articles", "robotics_publications"),
   },
+  {
+    title: "AI safety publications",
+    key: "ai_safety_publications",
+    aggregateType: "median",
+    category: "publications",
+    ...generateSliderColDef("articles", "ai_safety_publications"),
+  },
+  {
+    title: "Large language model publications",
+    key: "llm_publications",
+    aggregateType: "median",
+    category: "publications",
+    ...generateSliderColDef("articles", "llm_publications"),
+  },
   {
     title: "All publications",
     key: "all_pubs",