I don't think that a clear dividing line will be reached between human tools, robots, and higher-intelligence automata. This is because while these are augmentations of baseline task capacity, they are not significant extensions of tasks conception, and have little grounds to be.
And yet, this lack of clear escape of machines from human task conceptions, helps clarify perhaps the machine domain vs the human domain: machines are limited to tasks and humans are not.
A physical stick to provide reach or leverage, or a pot to gather materials or keep something out, are the most basic material tools, and we understand them as augmenting task capacity.
Beyond passive, static tools, arise mechanical tools, with articulations, processes and automotivation in the form of stored-and-released energy. Archimedes screws, trebuchets, aquaducts (converting potential into kinetic energy), and onwards to canons, steam trains, and beyond.
Again, these exhibit extensions to task capacities, far beyond simple localized tool support to individuals, but clearly an advanced of them also.
When we come to machines which fully encapsulate and expand task capacity directly, such as a digger, we can explore more clearly the evolution pathway of augmentation technology.
A digger is responding to instructions from a human, to perform a task the human is directly supervising, in real time. Augmentation is the natural word to describe this, since the task, and each step within it, has analogues at earlier and simpler stages of tool development.
What happens, though, when we look in the other direction? What additional augmentations will come to be?
In the first instance, in addition to simple scale and speed enhancements, which are obviously the classical domain of machines, a modern digger may add capabilities in terms of precision, even error-correction. For example, if the user is manipulating the digger in such a way that the bucket is vibrating and spilling contents, a mechanical or electrical damping feature might be applied to reduce spillage.
Beyond this a feature is imaginable of simply repeating a task sequence previously directed by a human - such as a move - lift - move - drop procedure, with the human in direct supervision. And then the same task, without supervision - as a true automation.
Where robots take over from this increasingly extending automation track is the diversity of automated tasks they can perform, and the level of abstraction at which the tasks can be defined.
For example, if a robot - autonomous - car is asked to drive a certain route, it might have the autonomous capacity to do this in various types of weather, with different road conditions. But if it asked to simply find the quickest route to a certain destination, the task has been set at a higher level of abstraction: it's not merely the repetition of a previously defined task.
So human task capacity is moving along a line of technical sophistication, autonomous capability and power.
But what about the conceptual capacity of the automation? What is the evolution in task conception?
I think asking this question clearly reveals that material and technical capacity of tools/machines/automata is extending along a very narrow pathway of task conception - and indeed space of potential task conceptions may be getting narrower as the field of task capacities expands.
In the progression of task capacities described above, what is the evolution fo the task concept itself? Apparently, none. In all versions of the task capacity, the nature of the task is to perform something the human would have wanted to do themselves, or at least achieve an outcome that the human has previously conceived and would like to have accomplished.
It's possible to conceive of fantastical tasks, at the outer boundary of human imagination, and plan how machines might do these - say terraforming other planets at scale, or Dyson Sphere's to capture all the energy from a star, and so forth.
But past a certain point, these extensions of tasks concept, start feeling artificial: what are they for, in any detailed sense? Are they tasks or just operations? And they certainly feel limited - if they have any purpose at all - to the extent they sound like material or resource gathering exercises. This suggests that they aren't really extensions the task conception all that much anyway, they are just complex extensions of tasks capacity.
There are two arising considerations here.
First, tools, machine and automata, even as their task capacity rises to previously unimagined extremes, seem to be trapped inside a bounded tasks conception space.
Second, in observing this, we may have come to a natural boundary of the machine domain. And that boundary is not simply a limit in the nature of the tasks that machines and automata can engage in, but that the task - a purposive action with inputs, resources, signals, processes, outputs, success-and-failure criteria - may is the only domain they may operate in at all.
To put these two points together, it maybe that machines and automata not only cannot expand the task conception space meaningful, but that they cannot 'be' or 'do' anything other than instantiate tasks. This may be the limit of the machine domain: the task.
This is instructive, since it enables us to look back at humans and ask if such a limit applies to the human realm. And clearly, it does not, in so far as humans understand themselves in a vernacular or technical sense.
It's possible to reverse-engineer the functioning of the body in terms of 'task orientation': in the sense that the it is the 'task' of the heart to beat, and bone marrow to produce red blood cells. There is a danger with this approach however, since as we do this, we start applying a teleology into more and more abstracted material domains: such that it becomes the 'tasks' of atoms to combine as chemicals, releasing energy, and at the grandest level, it is the 'task' of the universe to produce entropy.
But questioning injection of teleology into the material functioning of sub-conscious dimensions of the human realm is not necessary to demonstrate that the human realm has more going on than mere task orientation.
Love, understanding, appreciation, spontaneity, creativity are dimensions of the human realm that seem to defy a task framing. What task is being performed in the act of understanding something, or loving someone?
This consideration, that the human experience goes beyond task orientation, and the fact machines are trapped inside a task orientation paradigm, and that their task concept seems very limited despite their expansion of task capacity, seems helpful in defining the machine realm, and thus the human realm.