经济学人双语精读：大语言模型越来越大，越来越好

2024年8月1日 文摘

相见恨晚的英语微信公众号：田间小站

AI’s next top model
AI的下一个顶级模型

Large language models are getting bigger and better
大语言模型越来越大，越来越好

Can they keep improving forever?
它们能永远进步吗？

IN AI-LAND, technologies move from remarkable to old hat at the speed of light. Only 18 months ago the release of ChatGPT, OpenAI’s chatbot, launched an AI frenzy. Today its powers have become commonplace. Several firms (such as Anthropic, Google and Meta) have since unveiled versions of their own models (Claude, Gemini and Llama), improving upon ChatGPT in a variety of ways.
在AI世界里，技术从举世瞩目到老套过气的速度堪比光速。不过18个月前，OpenAI的聊天机器人ChatGPT的发布掀起了一场AI狂潮。今天，它的本事已经变得司空见惯。几家公司（如Anthropic、谷歌和Meta）继OpenAI之后推出了自己的这类模型（Claude、Gemini和Llama），在ChatGPT的基础上朝着不同方向提升。

  • frenzy 表示“疯狂、狂乱、狂暴、狂热”，常用搭配 frenzy of sth 表示兴奋或狂野行为控制不住的状态或时期（参见：小词详解 | frenzy）
  • commonplace 表示“常见的事物、平常的事物”以及贬义指“平庸的东西”，常作正式用语使用，也就是经常做或发生的事物，以至于不再特别、无趣或没有挑战性（参见：单词双拼 | commonplace 并非什么寻常地方）

That hunger for the new has only accelerated. In March Anthropic launched Claude 3, which bested the previous top models from OpenAI and Google on various leaderboards. On April 9th OpenAI reclaimed the crown (on some measures) by tweaking its model. On April 18th Meta released Llama 3, which early results suggest is the most capable open model to date. OpenAI is likely to make a splash sometime this year when it releases GPT-5, which may have capabilities beyond any current large language model (LLM). If the rumours are to be believed, the next generation of models will be even more remarkable—able to perform multi-step tasks, for instance, rather than merely responding to prompts, or analysing complex questions carefully instead of blurting out the first algorithmically available answer.
这种对新模型的渴望不断加速。3月，Anthropic推出了Claude 3，在各种排行榜上击败了之前OpenAI和谷歌开发的顶级模型。4月9日，OpenAI通过调整其模型夺回了（以某些指标衡量的）桂冠。4月18日，Meta发布了Llama 3，初步结果表明它是迄今为止功能最强大的开源模型。OpenAI很可能会在今年某个时候制造轰动，它可能发布能力超过当前任何大语言模型的GPT-5。如果传言可信，那么下一代模型还要更加惊人——例如，能够执行多步骤任务，而不仅仅是就提示给出回应，或者能仔细分析复杂的问题，而不是直接给出通过算法得出的第一个答案。