Idea Learning combines the power of human knowledge with machine learning.

See how Gamalon works.

Request A Demo


Gamalon gets smarter faster by bringing together machine learning with domain-specific human understanding.

Gamalon was built to be explainable and easily editable.

Idea Learning is the fastest, most accurate approach to help you gain rich insights from your unstructured data.

  1. Import Real-time Data

    You can import your real-time or batch data from any other system. Idea Learning incorporates human grammar automatically, and it rapidly learns to understand data in any database or machine format.

  2. Our AI Summarizes your Customers’ Ideas

    After importing data, Idea Learning finds the initial set of categories to populate your machine learning model and surfaces a first summary of all the customer’s ideas.

  3. A Person Updates the Model

    A person, typically a subject matter expert for the business, reviews what the machine learning has come up with in terms of categories and so forth, and provides any feedback/guidance they may have for what they would prefer.

  4. Our AI Updates the Model

    After an SME or analyst is done editing the model, our Idea Learning algorithm will then iterate again on the data set, learning more, and the model will grow deeper.

6 Key

We have developed our approach to machine learning, which we call Idea Learning, with 6 key principles in mind. Inspired by the work of Pedro Domingos in the Master Algorithm, we combined different aspects of each approach to machine learning.

  • 1. People

    Editable & Auditable AI; Human-machine interaction using ideas, not just labeling data
  • 2. Deep Learning

    Neural / Connectionist; Deep learning finds patterns that people cannot find themselves.
  • 3. Programs

    The ability to learn rules/programs that humans can read and understand
  • 4. Probabilistic

    System is self aware internally about where it is uncertain
  • 5. Populations

    Inspiration from evolutionary algorithms
  • 6. Compositional

    Train and then recombine modules