Question

In column 695 it describes how AI works, and that was correct until a few months ago. Today it seems that AI machines work in the form of MoE and definitely imitate processes of drawing conclusions, and not just responding to input-output. See how DeepSeek v3 works. Its output also includes its “thinking” process, which is amazing, because it shows how it breaks down the problem. From day to day the machines are learning to analyze (I’m not writing “to think,” because that’s exactly the discussion) in a deeper and deeper way what exactly was asked and how to construct the answer… I’d be happy if the Rabbi would address this as well.