2024 has been a year of remarkable progress in AI. Over the past year, I’ve grown from a beginner to a deeply engaged specialist, fully immersed in the field’s rapid advancements.
Things have been evolving at an incredible pace, and conversations about AGI and ASI seem endless.
I recently came across an X post that perfectly captures my feelings—it resonates deeply with my experience this year.
Here's the table of contents for my end-of-year review of things we learned out about LLMs in 2024 - we learned a LOT pic.twitter.com/VbLhw7scF4
— Simon Willison (@simonw) December 31, 2024
In case you are my Chinese funs, here is the Claude 3.5 sonnet translation：
- GPT-4的技术壁垒被全面突破
- 一些GPT-4模型可以在我的笔记本电脑上运行
- 由于竞争和效率提升，大语言模型价格暴跌
- 多模态视觉变得普及，音频和视频功能开始浮现
- 语音和实时摄像头模式从科幻变为现实
- 提示词驱动的应用程序生成已成为一种商品
- 对最佳模型的普遍访问仅持续了短短几个月
- "智能代理"还未真正实现
- 评估真的很重要
- Apple Intelligence表现不佳，但Apple的MLX库很出色
- 推理能力扩展模型的崛起
- 目前最好的大语言模型是否在中国以不到600万美元的成本训练完成？
- 环境影响有所改善
- 环境影响变得更加严重
- 草率/潦草之年
- 合成训练数据效果很好
- 大语言模型的使用变得更加困难
- 知识分布极不均衡
- 大语言模型需要更好的批评/评价机制
- 2024年我博客上所有标签为"llms"的内容
I’ve also put together my personal summary of AI advancements in 2024, documented as video talk.