The popular understanding of the term 'evolution' may be correct, but it is not the same as the biological term.
Stephen Jay Gould would have it that we hold in our minds two different sets of concepts when we use the term evolution.
The original use, and the prevailing one in general use, is as a synonym for predictable progress to an inevitable end. Gould illustrates this with quotes from the Encyclopaedia Britannica, discussing star formation. In describing the process, the term evolution is used as if the process, once it kicks into gear, is entirely deterministic - that is, given one set of input factors, the outcome is pretty much pre-ordained. And this is correct; and such determinism underpinned my years of physics training - notwithstanding different rules applying at the extremes, such as quantum levels.
According to Gould, Darwin largely shied away from use of the term, because of this determinism. There are two problems here: biological evolution lacks specific direction and predictability. Yes, it is non-deterministic.
Direction: fit to environment is the key. We can be misled into thinking that we are the pinnacle of evolution, and so all evolution is towards increasing complexity. Gould discusses a type of parasite called dicyemids, that live in the renal organs of squids and octopi. There had been much debate over whether these creatures had always been primitive, or whether they had shed functionality to adapt to their environment, simplifying in the process to little more than feeding and reproduction functions. In 1999, Kobayashi, Furuya amd Holland presented genetic evidence in Nature magazine to demonstrate dicyemids had in fact descended from more complex creatures, in the process becoming incredibly simplified. Evolution is not necessarily about increasing complexity. Progress can be in any direction appropriate to the needs of the environment.
Predictability: Darwin's "descent with modification" occurred, as Gregor Mendel demonstrated in the 19th century, via random genetic mutation. In a static environment, a species' genetic outliers are disfavoured; in a changing environment, those random outliers can be better equipped to handle the changes. Gould's example here is an elephant in Siberia: when the climate turns cold, those elephants in successive generations that are hairier are more favoured to survive.
However, this randomness is such that, were history to be replayed, there is no guarantee that the changes wouldn't play out in a different way. There is no promise of predictability, since results are entirely dependent on what mutations eventuate - and whether the bearers of the altered genes survive whatever mishaps and chance comes their way.
All this is not to deny that increasing complexity has at times been a definite trend in the planet's history. Increased complexity be a survival advantage. Further, evolution has to a significant extent been an "arms race" between predator and prey, between competitors for resources. Arsenal improvements, including lung capacity, musculature, size (in both directions) and functionality have always aided genetic longevity.
It's barely worth trying to eliminate concepts of determinism and direction from popular understanding of the word 'evolution'. But it suffices to understand the different uses in the different contexts.
Gould, Stephen Jay (2002): What Does The Dreaded "E" Word Mean Anyway? in I Have Landed, Jonathan Cape, London.