Experiment with new algorithms and frameworks, and think about where those will fundamentally change business in the near term.
Which new algorithmic capabilities and frameworks are about to change the way business is done? Do new ideas presented in research papers translate to advantages for real-life data sets? Are new libraries mature enough to be used in production? The only way to truly find out is to experiment, and build a potential product yourself. Fast Forward Labs, a machine intelligence startup in Brooklyn acquired by Cloudera, does exactly that and reports on the "recently possible".
We worked on various topics which were published in the form of reports and accompanying prototypes, demonstrating the capabilities of artificial intelligence. For example, we reviewed the of state of art for labeling training data with active learning, and we experimented extensively with transfer learning for neural networks, which can be used in natural language processing workflows.