Nowadays Proof of Concepts are the way to go to start with AI. But with limited budgets and optimistic expectations, a majority of POC’s do not get to the next stage. Our experience at Hemisphere is that setting the right expectations and looking beyond POC’s is the way to go.
Setting expectations on a technology that is not fully understood is not an easy task. With limited in-depth knowledge of AI it is hard to determine the possibilities. Overestimation is a common problem when it comes to setting targets for AI projects. We see many great examples based on big and ‘perfect’ datasets. But in many cases this does not hold. Before starting/continuing with AI it is key to evaluate the project according to the difficulty of the task and the required performance level as can be seen below.
Read more here
Written by Lars Rigter