Citing software like autoland which slavishly applies the rules humans have programmed into it as a reason why neural networks should be considered better than humans at solving general intelligence problems is moving the goalposts even further the other way...