3 ms·
Worse! It's trained to output coherent reasoning, so by putting the assumption last there's a risk it massages the assumption slightly to fit the conclusions it
by kqr 8mo ago
Worse! It's trained to output coherent reasoning, so by putting the assumption last there's a risk it massages the assumption slightly to fit the conclusions it has already drawn.
- cracki 8mo agoThat's a real danger, yes. If it's the reasoning kind, then it'll run through one iteration in the background before it composes its emissions for the meatbag.