I felt that way maybe an year or two ago. It seemed like the most research were only concerned about building bigger models to beat benchmarks. There was also t

I felt that way maybe an year or two ago. It seemed like the most research were only concerned about building bigger models to beat benchmarks. There was also this prevalent idea that models need to be big and have massive compute. Especially from companies like openai. I was glad that models like deepseek were made. Bought back some hope