Looks like I might be the minority, but I disagree with this prediction. Better models will also be better at abstracting and we have seen several examples (e.

Looks like I might be the minority, but I disagree with this prediction. Better models will also be better at abstracting and we have seen several examples (e.g. the paper LIMO: Less is More for Reasoning) that with a small amount of training data, models can outperform larger models.