When a client asked us recently to explore the potential for a machine-learning platform, we took the tried-&-true Igniter approach: after a toe-in-the-water investigation we planned some prototyping and tests. The first of three steps involved a bunch of no-holds-barred experimentation. How might such a system develop and apply its own learning, autonomously? To cut a long story short, the results pleased and surprised us all. They also allowed us to address two major questions in advance:
What system(s) would the AI operate in during its initial prototyping, and;
How would the learning and prediction specifications act?
Once the client's risk had been mitigated, we knew how to proceed. More importantly, by the time major investment was needed we were sure of success.