Why Google is betting on RSI, starting with Gemini 4 Argon
On Sept 30 Google released Gemini 4 Argon, the first model in its Gemini 4 series, the same day a Google DeepMind researcher posted that the model had "surpassed" him. The piece traces Google's September RSI steps — Gemini 3.8 Flash, Dream-RSI and RRSI — plus AlphaEvolve's use inside data centres, chips and AI training, arguing the company wants AI to help build its next models. Co-founder Sergey Brin is reported to be steering more resources toward RSI, as Alphabet raised its 2026 capex guidance to $195–205 billion.
If AI can keep improving training code, toolchains and infrastructure, leadership may hinge on how fast a lab iterates rather than on any single model generation — though RSI still lacks a common definition and accepted way to verify gains.
“我们的模型正式超越了我。我已经没什么可教它的了。”
9月30日，Google DeepMind研究员Zirui Wang在社交媒体上写道。
当天，谷歌发布Gemini 4系列首款模型Gemini 4 Argon。这是谷歌今年最重要的一次旗舰发布。
这句半开玩笑的感叹，很快引出一个更大胆的猜测：谷歌是否已经实现了RSI？
RSI，递归式自我改进，是指AI参与改进自身或研发下一代AI。
如今，多家公司已将RSI写入研究目标和投资叙事。相关论文、评测和产品也在密集出现。
#Google#Google DeepMind#Gemini 4 Argon#Gemini 3.8 Flash#AlphaEvolve#Dream-RSI#RRSI#OpenAI#Anthropic#Claude Code#Scale AI#Alphabet