In machine learning, to train a neural model, one typically needs a lot of data. This is challenging for many clients, as access to that data isn't always easy- this is when transfer learning comes in handy. Transfer Learning: Reusing a previously trained model on a new problem, especially when the new problem has a similar structure or features to that of the target domain. This is particularly valuable in the field of data science, as most real-world situations do not require millions of labeled data points to train complicated models. Blinx AI's transfer learning feature is one of the best ways to speed up and reduce the cost of training your AI model.

I was reading two articles this week on MIT Technology Review about the difficulties of explaining the decision-making of advanced algorithms that uses AI. This explanation is fundamental as our life's become more intertwined in ways that sometimes we do not even realized. From self-driving cars, who's approved for a loan, and personalized medicine the issue of a methodology to explain the outputs of AI is becoming to the forefront due to liability issues. After 20 years implementing advanced algorithms filed and 12 years in the legal profession I have come with a methodology that help explaining deep learning results in such a way that is understood in layman's terms. Below is how I would create a minimum viable product (MVP) to develop an application to explain AI outputs.