200k now, reasonably speaking a few million is within reach, which is reptile/fish range, the terrifying thing is though that if they train this to imitate humans (which they will) who knows how many orders of magnitude of efficiency gains you get (in terms of neurons needed for a certain level of consciousness) versus natural organisms that are dependent on natural evolution and need to support other bodily functions basically irrelevant to consciousness.
It seems unlikely that we would be more efficient at achieve consensus than evolution which can hand craft neural structures via feedback loops across millions of generations.
Especially when this demo needs 200k neurons when organizations with vastly fewer neurons have more complex behaviors.
We already know we can be more efficient than evolution at many tasks. Pelicans after all never developed jet turbines. We may not be able to access a simulation space as vast as evolution does but for small solution spaces we do quite well.
The problem with that logic is that evolution iteratively builds on top of old systems. The foundations are often remarkably crufty.
My favorite concrete example is "unusual" amino acids. Quite a few with remarkably useful properties have been demonstrated in the lab. For example, artificial proteins exhibiting strength on par with cement. But almost certainly no living organism could ever evolve them naturally because doing so would require reworking large portions of the abstract system that underpins DNA, RNA, and protein synthesis. Effectively they appear to lie firmly outside the solution space accessible from the local region that we find ourselves in.
I agree with your second point though that this system is massively more complex than necessary for the behavior demonstrated.