随着公共和合成数据集逐渐耗尽，人工智能实验室正转向购买已倒闭初创公司的代码和工作区数据以训练模型。1 这一趋势源于业界对真实世界记录需求的增加，旨在更好地训练AI代理（AI agents）。1 然而，收购失败初创公司的数据引发了法律与伦理层面的争议。1 例如，加密货币创业者Keelan Miskell在关闭其初创公司后，通过出售数字档案获得了数万美元收入。1
除AI数据获取外，科技领域还出现了利用人工智能解码动物通讯以实现跨物种交流的进展。1 与此同时，太空探索与气候行动也面临新的挑战与目标：NASA希望在本世纪末建立小型月球城市，且美国与中国正在竞争建设月球基地。1 在环保方面，Apple曾于2020年设定目标在本十年末实现净零排放，而中国科学家开发出能将火星大气中的二氧化碳转化为液态甲醇的催化剂。1 此外，Cloudflare首席执行官Matthew Prince指出，未来两年平均每周会出现一个关键的互联网漏洞。1
As public and synthetic datasets become exhausted, artificial intelligence laboratories are increasingly purchasing codebases and workspace data from failed startups to train their models 1. This trend reflects a growing demand among AI developers for authentic real-world records to build effective AI agents 1. The practice has sparked significant legal and ethical debates regarding the ownership and use of such digital assets 1.
Keelan Miskell exemplifies this new market dynamic by generating tens of thousands of dollars in revenue after selling his digital archives following the closure of his cryptocurrency startup 1. Beyond the acquisition of proprietary data, other major technological developments include China's creation of a catalyst capable of converting Martian atmospheric carbon dioxide into liquid methanol 1, while Matthew Prince, CEO of Cloudflare, warns that a critical internet vulnerability is expected to emerge every week over the next two years 1.
评论
还没有评论，欢迎留下第一条。