Keep on to blur preview images; turn off to show them clearly
RT @getpy: If you are in Bengaluru and want to attend DSPy meet-up it's happening on 20th Saturday. 19 more seats remaining. Please RT #DS…
Asst professor @MIT EECS & CSAIL (@nlp_mit). Author of https://t.co/VgyLxl0oa1 and https://t.co/ZZaSzaRaZ7 (@DSPyOSS). Prev: CS PhD @StanfordNLP. Research @Databricks.
Specifically, I think that the US frontier has much better post-training coverage of tasks, specifically tasks similar to these ones, except for two. So you bet a puncture in the “usemaxing advantage” balloon. Base intelligence is of little relevance here. @TheZvi
We're in a race. It's not USA vs China but humans and AGIs vs ape power centralization. @deepseek_ai stan #1, 2023–Deep Time «C’est la guerre.» ®1
ChatGPT “成人模式” 即将上线 OpenAI 计划在 2026年第一季度（Q1 2026） 推出 ChatGPT 的 “Adult Mode（成人模式）”。 OpenAI将引入年龄识别机制，自动保护未成年人不接触敏感内容。 详细内容：https://t.co/WsoEbc1Ke5
带你了解全球最前沿科技、AI动态... 学AI找小互，找小互，上 https://t.co/4PVaHEr5r3 ... 小互AI日报 社群：https://t.co/LIEXfWUHv1
> also achieves a new best individual score on 2/17 tasks, most notably beating the SOTA on chess-winners (which previously looked like it had a hard ceiling) by > 5 pp So yeah. Speciale is not so Speciale when beaten by oss-120b. But as I’ve said, its *method* is now frontier.
Specifically, I think that the US frontier has much better post-training coverage of tasks, specifically tasks similar to these ones, except for two. So you bet a puncture in the “usemaxing advantage” balloon. Base intelligence is of little relevance here. @TheZvi
更少错误，更高稳定性、更懂人心 GPT-5.2 的“幻觉率”（错误回答率）降低 38%。 它更可靠地回答研究、写作、分析类问题， 减少了“编造事实”的情况。 同时在心理健康相关任务中，模型回应的安全性显著提升。 在心理健康、自残、自杀与情绪依赖等敏感场景中表现更稳健。
ChatGPT “成人模式” 即将上线 OpenAI 计划在 2026年第一季度（Q1 2026） 推出 ChatGPT 的 “Adult Mode（成人模式）”。 OpenAI将引入年龄识别机制，自动保护未成年人不接触敏感内容。 详细内容：https://t.co/WsoEbc1Ke5
Coding couldn’t be more fun right now. I feel like this is my reward for writing so much code the barbaric way for so long. Thank god I never realized how much faster things could be.
Opening portals to VR without headsets at @portalvr_io. Problems soluble, potential to improve invariant.