It's better that way
Disregarding how this is implemented biologically, the answer to why is because it provides optimal results.
In learning systems where you are able to freely adjust the ability to learn - e.g., AI research 'reinforcement learning' paradigm, there is a tradeoff commonly called Exploration vs Exploitation - i.e., should the agent focus on gaining new experiences by trying out new solutions that might be worse or better (but you don't know), or should it do things in the way that it believes to be best, and ignore the other options. Quite intuitively, you do get the best results by focusing on exploration early-on when you have little knowledge, and avoiding exploration after you have 'sufficient' amount of experience.
Similarly, in learning systems which "passively" receive new knowledge, there is a balance decision - how much should you alter your beliefs after seeing new evidence. If a two year old kid saw a flying man, it'd make sense to believe that men could fly - after all, two year olds see radically new phenomena almost every day, and it is useful to adjust their beliefs to note "ok, X is working that way, and Y always happens before Z" - even if they'd just seen it a single time. However, if a sixty year old saw a flying man, it'd instead make sense to look for wires and a movie camera crew. Similarly, in machine learning systems it is optimal to slow down belief adjustment(=learning) with time - as you've accumulated a larger baggage of earlier evidence, the newer evidence is comparatively much less important.
Now, going back to people - as per the above examples, in any learning system, you get the best results by slowing down your learning rate with time. Learning is a very useful capability for survival&reproduction; doing learning properly is a competitive advantage in natural selection.
So there is evolutionary pressure to do it this way - it should be expected that all animals, including us, would have evolved some biological way to adjust their learning so that it slows down with age. Animals would experiment and learn during adolescence, and afterwards mostly stick to the already formed habits. Changing neuronal plasticity is one way to implement this, there may be others.
Of course, it may be that it's no longer optimal given the radical change in our lifespan - if you live twice longer, then it'd be useful to learn twice longer as well - but it's originally implemented because it is (was?) actually an advantage to have it that way.