一位来自谷歌软件工程(SWE)岗位的开发者近期展示了一项实验,通过部署48个AI代理并行处理约1,200平方公里的NASA激光雷达(Lidar)数据图像,在不到72小时的时间内识别出一处疑似未被登记的玛雅遗址 1。该过程始于10月18日,至10月21日开发者已收到墨西哥政府机构的官方回复,整个从构思到获得反馈的流程极为迅速 1。被识别的地点位于墨西哥境内,初步判断为古典时期(约公元250年至900年)的小型玛雅行政中心,其结构特征包含金字塔、广场及水池 1。
参与此次数据分析的AI模型涵盖了Gemini、ChatGPT、Claude和Grok等多种主流大语言模型 1。尽管该发现在初期被视为新发现的考古成果,但根据11月11日的更新信息确认,该遗址实际上已在Estrada-Belli等人于2025年发表的研究中被测绘过,只是未录入墨西哥国家人类学与历史研究所(INAH)的官方登记册 1。因此,这一事件更准确地定义为对该遗址的“重新发现”,而非全新的考古突破 1。尽管如此,这一案例有力证明了通用AI代理在处理非编程领域复杂科学数据分析任务时的潜力与效率 1。
A developer from Google SWE utilized a swarm of 48 artificial intelligence agents to analyze NASA lidar data, resulting in the identification of an unregistered Maya archaeological site in Mexico 1. The entire process, from the initial concept to receiving an official response from government authorities, was completed in under 72 hours between October 18 and October 21 1. The AI models employed for this task included Gemini, ChatGPT, Claude, and Grok 1.
The parallel processing system examined approximately 1,200 square kilometers of lidar imagery, uncovering what appears to be a small administrative center dating back to the Classic Period (roughly 250–900 AD) 1. The discovered structure includes a pyramid, plazas, and reservoirs, though it had not been logged in the registry of Mexico’s National Institute of Anthropology and History (INAH) 1. While the find initially suggested a new discovery, an update on November 11 clarified that the site had actually been mapped in previous research by Estrada-Belli et al. in 2025, characterizing the event as a "rediscovery" rather than a completely novel finding 1.
评论
还没有评论,欢迎留下第一条。