For robots to serve us better, they need to be able to tell the difference between humans and other, potentially malicious robots.
Every time we prove we're not a robot through reCAPTCHA, robots get a tiny bit smarter. This means that in the future, the threshold for proving that we're not robots will go up, which means that we'll have to answer more complex questions to the robots that ask us for the proof that we're not robots. The increasing complexity of those questions answered also increases the quality of the learning potential for the robots.
It's an interesting cycle that just got started in the last years
In the future, we'll have gained a lot of time to spend, having to work less because robots will work for us.
But we'll spend more time proving that we're not robots in return.
Right now, this development is marginal, but in a few years, with exponentially increasing processing power and decentralized machine learning, it'll become significant.