AI isn’t just one model, its many models that all interact

Years of experience building machine learning systems has taught us a key theme — there is more to the problem than the “train, validate, deploy” verbiage we see in many competitive offerings.

What is the value of your data? How can it be leveraged in ways that solves business objectives? How can the outputs of one model trigger actions in other models?

TensorML is based on the hypothesis that real Machine Learning systems should function like a collective, emergent system. One system might read text scraped from the web, index and extract relationships like entities or funding events, while scoring relevance. Other systems can then use those outputs as inputs that trigger incremental/transfer learning systems, regress those outputs on, say, user behavioral data, and then write custom marketing messages, or recommend new products that are locally, contextually relevant. Or make bids in stock markets or ad networks, write a custom tweet, or show emergent patterns.

Understanding how bespoke modeling systems can interact, respond and adapt to data is the new digital gold.