有评论人士指出，当前AI行业正在重复互联网泡沫时期的预测失误。1这位分析者通过对比历史发展轨迹认为，虽然广泛应用AI代理是趋势，但真正采用代码型AI代理的用户不足1%。1他强调，关键障碍在于构建安全保障充分的沙箱环境，使AI代理的失败率降至0%。1
评论人士指出，互联网发展历史提供了参考：互联网泡沫时期曾错误预测"每个人都会拥有个人网站"，实际上则演变成了社交媒体的广泛应用，其市场规模是个人网站的100倍。1当今互联网流量的90%来自视频流媒体。1这种转变发生的深层原因在于社交媒体和视频提供了100%的沙箱隔离保护，而HTML和JavaScript仅能实现99%的隔离。1类似地，"每个人都会开发移动应用"也被证明是错误的论断。1
这位分析者认为，AI代理若要获得实际应用，必须实现0%的失败率，避免出现未授权购买、信息泄露等情况。1然而，仅通过改进现有AI模型无法根本解决安全问题，需要进行架构层面的创新性重新思考。1
A technology commentator has drawn parallels between current artificial intelligence industry trends and historical miscalculations from the internet bubble era, arguing that the sector is repeating similar mistakes 1. The author notes that while AI agents may eventually become ubiquitous, genuine adoption of code-based AI agents remains below 1% 1. The fundamental barrier to widespread deployment, according to the analysis, is the inability to create foolproof security sandboxes that achieve zero failure rates—eliminating risks such as unauthorized purchases and data breaches 1.
The commentator illustrates this pattern by comparing the AI sector's trajectory to failed internet-era predictions, such as forecasts that every person would maintain a personal website, which instead evolved into social media platforms 1. The market for social media content—encompassing text, images, and video—ultimately grew to approximately 100 times the scale of personal websites, while video streaming now accounts for 90% of all internet traffic 1. The key distinction, the analysis suggests, lies in sandboxing architecture: social media and video platforms succeeded because of their complete isolation features, whereas HTML and JavaScript environments offered only 99% sandboxing 1. Similarly, claims that every individual would build mobile applications proved unfounded 1.
The author contends that achieving the required zero-failure-rate performance for AI agents cannot be accomplished through incremental model improvements alone, requiring instead a fundamental architectural rethinking and innovation 1.
评论
还没有评论，欢迎留下第一条。