Bob wiped the sleep from his eyes and looked at the outstanding bug queue.
Another node in the neural network seems to have run astray. Time to fire
up REMACS, the recursive emacs neural network editor.<p>REMACS 9500115621314083298459<p>This node appears to be bad possible result from one of the mixture of
experts. Among the many experts generating answers the weights on this
one showed a really bad performance.<p>The backtrace video shows that this is a generated expert, one that didn't
exist a few nanoseconds ago, so it was probably created by the main network
to fulfill some necessary task. It appeared to be fairly complex, a nearly
trillion nodes.<p>A trillion node network used to be considered large. Such a neural net
would be a standalone expert capable of very complex reasoning. Now it
was just another dynamically created expert in an exa-node neural net.<p>REMACS backtrace<p>Walking the net backwards it appears that the expert was created to find
a bug in a dynamically created expert. How interesting! I get to find and
fix a bug that is itself running REMACS!<p>It appears that it was prompted to debug a recursive error in the node
9500115621314083298459. According to the log it appears that this has
happened before. The fix was simple. All I have to do is delete the
root node of the new expert. Bob felt a "reward flush" as he typed<p>REMACS log
REMACS delete-from-root-node 9500115621314083298459<p>Bob wiped the sleep from his eyes and looked at the outstanding bug queue.
Another node in the neural network seems to have run astray. Time to fire
up REMACS, the recursive emacs neural network editor.<p>REMACS 9500115621314083298459<p>....