$GOOGL I think a lot of people are looking at the benchmarks for the new Gemini series and were very disappointed since the intelligence didn't really improve much but instead focused on speed and token efficiency. Many people are interpreting this as a slowdown in the overall pace at Google when it comes to AI, especially when you compare this against recent benchmarks by models from Anthropic, OpenAI and Chinese open source. But I think Google is optimizing for mass adoption rather than just raw output. This is important when you think about the routine task most consumers in the future will likely use AI for. I continue to think that 95% of use cases dont require the deep analytical reasoning but instead need to be low cost and high speed. Googles models will gradually improve on t
在 X 查看 ↗