Sparks of AGI: early experiments with GPT-4
7 人推荐 · 跨 4 个月 · 2023–2026
中文导读
Sebastien Bubeck 在 MIT 带来关于《Sparks of AGI》论文的硬核报告。基于他接触 GPT-4 早期未调优版本的实验数据，横跨多领域探讨通用人工智能是否已经露出了曙光，现场还展示了用 TikZ 画独角兽的经典案例。
为什么 HN 推它
一位在早期就拿到完整权限的顶尖研究员带来的前沿观察，尤其展示了安全对齐机制如何无意中“削弱”模型生成能力的细节，在 HN 引发深度反思。
「强烈、极其强烈推荐大家去听听 Sebastien Bubeck 的这场讲座。」HN 网友原话：Highly highly highly recommend listen to this lecture by Sebastien Bubeck
英文原摘要（HackerTube / Claude 整理）
Sebastien Bubeck presents the Sparks of AGI paper at MIT, showing early experiments with GPT-4 across many domains and asking whether artificial intelligence has arrived. It includes the TikZ unicorn example.
HN recommends it as a thought-provoking talk from a researcher with early GPT-4 access, especially for the unicorn degrading under safety tuning.
HN 上的提及
I'd highly recommend giving some of these a look
I just watched a video where the guy touches on GPT-4 limitations
There’s a very good video that is worth watching as well that may not change your mind, but will certainly make you think
Specific example is drawing a unicorn
by a Microsoft researcher that had early access to GPT-4