As soon as I get a sentient, self-improving humanlike android in a few decades (circa 2030?), one of the first things I'll do is ask him to make more money for me. How? I'll leave that up to him/her. Regulations will hopefully prohibit robots from making money through illegal activity (steal, make/sell drugs, weapons, etc...) so whatever s/he decides to do will stay legal.
What I'll do is give the android $1.00 and ask him/her, "Through any legal means, please make $100 out of that $1 investment as soon as possible, and upload your progress to my PDA in real-time. Once you make $100, keep going, and reach $10,000 as soon as possible. After that, $1,000,000, then $100m, then $10b, then keep going as fast as you can. Also, once you reach the $10m threshold, please put 1% of your earnings into my bank account from that point forward, once per hour, on top of the hour."
Flops and Bytes
Since the android's computation speed will zoom at around 2 Exaflops (whereas one average human brain is ~20 Petaflops), the computing power will be more than enough to make huge profits out of tiny investments.
His/her memory storage capacity will amount somewhere in the ZETTABYTE range, so with a worldwide database on how to make money, in every last language, s/he will draw upon the most efficient ways to do so, and self-learn, self-adapt, and self-improve throughout the way.
It's reasonable to predict that 2 Exaflops will be commonplace by 2030. Here's a chart that graphs the Exponential Growth of Computing:
Even 20 Petaflops are sufficient, using full potential.
Even 20 Petaflops would be more than enough to make a huge fortune, but the economy progresses so comparatively slow because most humans do not use as much potential as the brain can muster. A non-biological android would do their very best to use as much of their potential as they possibly can.