AI Science Tracker

DolphinGemma: a model for finding patterns in dolphin sounds

Google; Georgia Tech, and Wild Dolphin Project · 14 Apr 2025 · biology

Discovery

Google; Georgia Tech, and Wild Dolphin Project

DolphinGemma: a model for finding patterns in dolphin sounds

Archived from discovery counts: the represented event introduces a model and planned field use, without establishing a new finding about dolphin communication. A model with about 400 million parameters, trained on decades of recordings of Atlantic spotted dolphins from the Wild Dolphin Project, predicts subsequent sounds to identify recurring patterns; it runs on Pixel phones used in the field. The patterns have not been confirmed to carry meaning.

Technology: DolphinGemma (audio model with about 400M parameters)

Publication: Announcement only · Test context: Test context unspecified in annotation. Catalog categories, not an evidence grade.

Result, AI contribution and limits

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

  • Publication status: Announcement only
  • Test type: not reported
  • Tested: Not reported
  • Still unknown: The source describes only early field testing with no quantitative metrics, the model is trained on one dolphin population, and the patterns are not confirmed to carry meaning.
  • AI role: A roughly 400M-parameter audio model predicted likely next dolphin sounds to find recurring patterns.

Evidence sources: https://blog.google/innovation-and-ai/products/dolphingemma/

The legacy label “Not reported” does not distinguish information absent from reviewed material from information not assessed. It is not a scientific finding of absence.

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: blog. Paper: not linked in this catalog; this does not mean no paper exists.

Review scope and date are not recorded; source checked does not establish full-text review.

Editorial assessment: Impact 2/5 · Validation 1/5 · Novelty 3/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Follow-up and relationships

No follow-up recorded; this is not a claim that no follow-up exists.

Follow-up review scope and date are not recorded.

No source-supported relationships recorded.

Browse the current catalog.