The article claims the difficulty is with the non-newtonian mortar, but this doesn't pass the sniff test. Non-newtonian fluids are easy to model and position feedback is common in virtually all modern automation systems. I say this as someone who used to do masonry work, worked in a rheology lab in college and is now an industrial automation engineer. The fact that numerous attempts to make bricklaying systems that didn't involve mortar but still did not succeed would certainly seem to indicate that mortar is not the dealbreaker. Indeed, the fact masonry assistant machines haven't seen wide adoption suggests that the problem has nothing to do with software.
Nature has optimized the design of muscles and skeletons over tbe course of hundreds of millions of years. People tend not to appreciate just how good of an actuator muscle is. You can certainly beat it in power, speed, precision, duty cycle, or form factor, but not all at once. Bricks and blocks are specifically optimized for being handled by a human laborer: they are small enough that a human arm can precisely manipulate them but big enough that you don't need to place an absurd number of them. Anything that's optimized for human manipulation will be difficult for a robot to manipulate.
A typical block weighs about 35 lbs. A robot arm capable of even lifting a payload that heavy will cost you somewhere on the order of 80k not including end of arm tooling, weigh hundreds of pounds, and draw several kilowatts of electricity. Then you need to take this arm and put it on some platform capable of moving around a job site such that it can reach any place a human could. At scale, you're probably looking at a quarter million dollar machine just to get a brick near where it needs to go. Combined with maintenance costs, it's likely such a system would never pay for itself over its useful life. Maybe you could do a little better with bricks but 5 lbs is still a lot for a precision robot. All the artificial intelligence in the world is useless if the actuators just aren't up to the task.
For better or worse, human muscle is cheap, it can be leased with no money down, and you typically don't need to pay for maintenance. That it comes equipped with an extremely powerful computing system for no additional charge is just icing on the cake. People tend to have the mistaken belief that things which are conceptually simple must be easy to automate and that things which are hard to automate must be secretly complex, but that's not really it at all. Tasks appear simple to us because we are good at doing them, and they appear complicated when we're not. Picking ripe fruit seems like such a simple task to us because that is exactly what we evolved to do. Simulating quantum chemistry, not so much. Picking fruit isn't secretly more complicated than quantum chemistry, but it's much easier to automate the latter because it's not hard to be better than a human at a task humans perform poorly to begin with.