Jared Kaplan (@Anthropic): AGI comes in 2027
Here is the translation:
Document
First, this issue is about a vision of the future. Azar and Kaplan discuss several different perspectives, with whom they agree that the future landscape may not be a single superintelligent entity, but rather a complex ecosystem.
This complex ecosystem may have many problems, such as AI interacting with each other and possibly producing unforeseen, harmful systemic consequences. Kaplan is worried that we may not fully understand these complex interactive systems.
They then turn to the macroeconomic and social impacts of this. Kaplan highlights a key issue: whether the boost in productivity from AI and its impact on the labor market will be faster and more intense than any previous major technological change? He emphasizes his basic stance, believing that AI is empowering at its core.
Finally, Kaplan mentions the importance of empirical research, saying that Anthropic is very concerned with understanding the effects of AI applications through actual data. They have developed an internal tool called Clio to summarize and analyze Claude’s usage, and are researching issues like whether AI complements or replaces existing work, and to what extent it increases productivity.
Kaplan also mentions changes in the software engineering field, which he thinks is a good starting point for research because software engineers themselves tend to be eager adopters of new technologies. Ultimately, Kaplan poses an unresolved question: Will the kind of transformation that has occurred in software engineering with AI also occur in all knowledge work? He believes that this answer may take time to explore.
In short, this interview discusses various issues such as complex ecosystems, economic and social impacts, empirical research, changes in the software engineering field, etc.
Translation
首先，这个问题是关于未来图景的展望。阿扎尔和卡普兰讨论了几个不同的视角，他们认为未来的图景可能不是一个单一的超智能体，而是一个复杂生态系统。这点他们都同意。
这个复杂生态系统可能会有很多问题，比如，AI 会相互作用，并且可能产生意想不到的、有害的系统性后果。卡普兰担心的是，我们可能无法完全理解这些复杂的交互系统。
然后，他们转向了对经济和社会的宏观影响。卡普兰提到了一个核心问题：AI 对生产力的提升和对劳动力市场的冲击是否会比历史上任何一次重大技术变革来得更快、更猛烈？他强调了他的基本立场，即认为 AI 从根本上来说是赋能的。
最后，卡普兰提到了实证研究的重要性，他说 Anthropic 非常关注通过实际数据来理解 AI 的应用效果。他们开发了一个内部工具 Clio 来汇总分析 Claude 的使用情况，并且正在研究像 AI 对现有工作是补充还是替代，以及它在多大程度上提高了生产力这些问题。
卡普兰还提到了软件工程领域中的变革，他认为这是一个理想的研究起点，因为软件工程师本身就乐于采用新的技术。最终，卡普兰问出一个悬而未决的问题：像如今 AI 在软件工程领域发生的变革会不会发生在所有的知识工作中？他认为这个问题的答案可能还需要时间去探索。
总之，这个访谈讨论了多方面的问题，如复杂生态系统、经济和社会的影响、实证研究、软件工程领域中的变革等。
Reference:
https://www.youtube.com/watch?v=tC_3KAyli4U