AI is now helping build better AI. So how close are we to it taking off on its own? 🚀 📄 𝗧𝗵𝗲 𝗘𝗰𝗼𝗻𝗼𝗺𝗶𝗰𝘀 𝗼𝗳 𝗥𝗲𝗰𝘂𝗿𝘀𝗶𝘃𝗲 𝗦𝗲𝗹𝗳‑𝗜𝗺𝗽𝗿𝗼𝘃𝗲𝗺𝗲𝗻𝘁 Tom Cunningham and others (METR, #Stanford, #MIT, #Yale, and #Columbia) 🔁 The idea is simple. Smarter AI helps build smarter AI,

AI is now helping build better AI. So how close are we to it taking off on its own? 🚀 📄 𝗧𝗵𝗲 𝗘𝗰𝗼𝗻𝗼𝗺𝗶𝗰𝘀 𝗼𝗳 𝗥𝗲𝗰𝘂𝗿𝘀𝗶𝘃𝗲 𝗦𝗲𝗹𝗳‑𝗜𝗺𝗽𝗿𝗼𝘃𝗲𝗺𝗲𝗻𝘁 Tom Cunningham and others (METR, #Stanford, #MIT, #Yale, and #Columbia) 🔁 The idea is simple. Smarter AI helps build smarter AI, which helps build even smarter AI. The big question is whether that loop can keep speeding up 𝘣𝘺 𝘪𝘵𝘴𝘦𝘭𝘧. 🧮 The authors turned it into one test. Each small jump in AI ability (think Claude 𝗢𝗽𝘂𝘀 𝟰.𝟳 𝘁𝗼 𝟰.𝟴) needs to make AI research 𝗮𝘁 𝗹𝗲𝗮𝘀𝘁 𝟭𝟱% 𝗺𝗼𝗿𝗲 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝗶𝘃𝗲. 📊 Where are we today? 𝗔𝗯𝗼𝘂𝘁 𝟵%, and that's probably generous. ✅ 𝗡𝗼𝘁 𝘁𝗵𝗲𝗿𝗲 𝘆𝗲𝘁. 𝘉𝘶𝘵 𝘵𝘩𝘦 𝘯𝘶𝘮𝘣𝘦𝘳 𝘪𝘴 𝘤𝘭𝘪𝘮𝘣𝘪𝘯𝘨. 🎯 The twist: AI could get brilliant at 𝘪𝘮𝘱𝘳𝘰𝘷𝘪𝘯𝘨 𝘈𝘐 without getting much better at 𝘵𝘩𝘦 𝘸𝘰𝘳𝘬 𝘵𝘩𝘢𝘵 𝘳𝘶𝘯𝘴 𝘵𝘩𝘦 𝘦𝘤𝘰𝘯𝘰𝘮𝘺. 🔍 Fair caveat: the data is thin, especially on how much AI really speeds up research, so treat these numbers as a first rough guess. Where do you think that 9% will be a year from now? 👇 #ArtificialIntelligence #AI #MachineLearning #AIResearch #FutureOfAI #Economics #AGI #RecursiveSelfImprovement #AISafety #TechTrends #Innovation #DeepLearning #GenerativeAI #LLM #Anthropic #OpenAI #METR #Productivity #Research #FutureOfWork Link to the paper: I started a monthly digest that cuts out all the noise in #Quant Finance & #AI/#LLM research. Just the things that actually matter. Friends keep telling me it saves them a ton of time. Sign up here: Here are past issues if you want a peek: