We keep giving machines mountains of data and then act surprised when they still fail basic reasoning. Large models can summarize entire libraries but miss a simple yes-or-no instruction. The problem is not the data. It is our belief that scale equals sense.
machine reasoning
1 post