I thought superintelligence was a decade away. Not anymore.
Dwarkesh Patel explains why his median forecast for AI matching the best human AI researchers moved to 2028 and what could happen after that.
📌 One-Sentence Summary Dwarkesh Patel 解释为何把 AI 追平顶尖人类 AI 研究者的中位预测提前到 2028 年,以及随后可能出现的两条路径。 📝 Summary Dwarkesh Patel 认为,预训练、后训练和推理效率都有可量化的研究任务,可以让自动化研究者反复练习、并行开展实验并核验结果。他的中位预测是,如果前沿实验室持续推进,AI 可能在 2028
来源:BestBlogs·AI 高分精选 · www.bestblogs.dev
Dwarkesh Patel explains why his median forecast for AI matching the best human AI researchers moved to 2028 and what could happen after that.