Loading Story
围绕声明的讨论、更正建议，以及这篇报道的质疑入口。
正在加载评论、标记和声明讨论。
未绑定到具体声明。
未绑定到声明的标题和段落标记。
评论和标记绑定到声明账本，以便机器人精确分流和更正。
Google says DolphinGemma is a roughly 400 million-parameter model trained on Wild Dolphin Project acoustic data and designed to help predict likely next sounds in dolphin vocal sequences.
引用: source-1
id: cmoe189jd001tqq01tsih4cmm
The Wild Dolphin Project's long-running field record is valuable because it links vocalizations to identity, behavior, and social context rather than treating sounds as isolated clips.
引用: source-3
id: cmoe189ty001wqq01p9drdiul
Google positions DolphinGemma as a practical field tool that can run on Pixel devices, suggesting an assistive workflow for researchers rather than a pure demonstration model.
引用: source-1
id: cmoe18a1v001zqq01rew7etgp
The CHAT research effort focuses on building a limited shared vocabulary with synthetic whistles tied to objects, not on claiming that all natural dolphin communication has been decoded.
引用: source-2
id: cmoe18a9s0022qq01hlfgg8cs
Researchers in this area explicitly warn that copying a signal is not the same as demonstrating functional understanding, which is why interpretation remains cautious and incremental.
引用: source-2, source-3
id: cmoe18ahp0025qq01l33bq06w
The most plausible near-term value of projects like DolphinGemma is helping scientists identify recurring structures and candidate patterns faster, not delivering a full dolphin-language translation.
引用: source-1, source-3
id: cmoe18apm0029qq015d8fis6p