[samegene321]

Yes and every halving of that 1% error is more than twice as hard to do with neural net training. At some point more principled control algorithms are needed.