A recent TechRepublic article “Now AI can write code for you” reported that Rice University researchers created an application called BAYOU that uses deep learning to write code for programmers. While the technology is in its infancy, it represents a major breakthrough in using artificial intelligence (AI) for programming software, and can potentially make coding much less time intensive for human developers. Would not it be interesting (and probably easier) to use a similar approach to building decision models for business people?
BAYOU trained itself by studying millions of lines of human-written Java code from GitHub, and draws on that to write its own code, according to the release. It is based on a method called neural sketch learning, which trains an artificial neural network to recognize high-level patterns in hundreds of thousands of Java programs. Bayou specifically generates lines of Java code that will work with available open APIs. In other words, a developer tells the system what they want to do, then the system finds examples of relevant APIs and provides code that may help them get what they want from those APIs. So far BAYOU is mostly useful for addressing one particular use case, but that use case is extremely valuable. CA Technologies published an article about this development under the roof of The Modern Software Factory Hub.
As a Decision Management Community we may start thinking about creation of an open “Decision Model Factory Hub”. No questions that a learning technique such as BAYOU will mature and be perfected by AI developers in a few years. To be applied to decision models (instead of software) it will need a publicly available repository of decision models for different business domains (instead of open APIs). There was an intention to create such a repository when we started the DMCommunity.org more than 4 years ago, but so far it hasn’t gone beyond several initial decision models and use cases – see https://dmcommunity.org/decision-models/. At the same time we saw quite active contribution of decision models to our challenges.
Now, 4 years later, the situation is different: we have an approved DMN standard supported by multiple vendors, the decision model interchange format, several graphical DMN modelers, and many real-world working decision models created by decision modeling practitioners worldwide. So, we need to start building a mechanism for creation of the Open Decision Model Repository, call it “Decision Model Hub” or suggest a better name. Such repository should be open as “open APIs” is and it shouldn’t be controlled by a commercial company. DM practitioners should be able to contribute their decision models to this DM repository with incentives and legal protections of doing that. Just having this repository available to the DM practitioners would be a great benefit even before the automatic (BAYOU-like) build of the decision models becomes a reality.
We ask our readers to share they thoughts how better to organize and maintain the open DM repository. We expect this will be among discussion topics during the upcoming DecisionCAMP-2018.