…The COVID-19 pandemic and accompanying policy measures caused economic disruption so stark that sophisticated statistical approaches were unnecessary for many questions. For example, unemployment jumped sharply in the early weeks of…
…Then, we trained this model using reinforcement learning on real programming tasks taken from actual Claude training runs, focusing on tasks that we knew were vulnerable to at least one reward hack…
…Since then, I’ve been trying to push AI towards more symbolic work (manipulating mathematical expressions rather than numerical data) and the core questions in theoretical physics. The hype There has been…
…Overall, these findings demonstrate that model predictions have meaningful correlation with real-world outcomes, at least in this domain, making them useful for comparing one task to another or tracking changes over…
…All the elevators are going to go crazy. This is just a philosophical question. These AI systems are fundamentally nondeterministic. They’re not predictable in the way that they operate, and…
…It’s leasing data centers that other companies are building on its behalf. In Abilene, Texas, that’s Crusoe, with whom Oracle has signed a 15-year lease; Oracle committed to paying…
…This was admirable and entrepreneurial, but it didn’t work—at least in terms of the bottom line. So we thought we’d do some hiring. First, we gave Claudius a manager…
…There's also xAI, which operates the chatbot Grok, perhaps best known for having guardrails so lax that it allowed users to generate sexual images of minors. And then there's X…
…Some of it Qwen covered better than I expected, or at least well enough that the gap didn’t really matter in practice. The thing I’m actually paying for has a…
…Or, at least, they're easy to game, and SWE-Bench Pro in particular has been criticized for containing existing pull requests that models with broad training data can essentially memorize. So…