Explainable AI is needed now; the time for experiments has gone

  • Stuart Battersby
  • April 28, 2020

Now, more than ever, AI projects need to deliver results quickly. As CMSWire notes, the times of long drawn out proof of concepts and AI experiments are gone; AI needs to deliver business results quickly.

Looking at this in the context of Explainable AI this is critical. Explainability is not a nice to have, it is essential to ensure fair, unbiased, ethical and transparent AI systems, and regulation is increasingly requiring explainability. This means that long proof of concepts and implementation timelines are not on the cards.

Chatterbox Labs’ patented Explainable AI software product is unique in explaining your existing AI assets without having to port all of your data to a new system (read more about that here). Our XAI software product sits as a layer, connecting to your AI model’s predict function. In fact, it is designed to explain AI models all across your business, irrespective of their underlying model or AI system.

In order to get you to production as quickly as possible, we have robust methods for connecting to your existing AI models. This is, after all, Enterprise software – we do not just dump a software library on you (leaving the implementation and programming up to you).

We operate on a connector basis (with prebuilt connectors to Google Cloud, AWS Sagemaker, IBM Watson & Microsoft Azure). In addition, we also connect to your existing, production ML models in your custom built stacks.

Deploying production ML models with JSON over REST is an industry standard approach (typically in a docker container). Alongside this API you will have an OpenAPI specification (formerly named a Swagger specification) which defines how your API operates (such as endpoints, formats, structures, authentication, etc). This specification is trivial to make (usually a few minutes) and is typically made as part of your deployment process.

Our XAI product can also connect to these systems in minutes, with no coding required. Using the user interface, simply point our software to your OpenAPI specification and follow a few simple steps to tell us about your API. Chatterbox Labs’ Explainable AI product will then auto generate the required connector with no additional input from you.

Of course, all your existing security controls remain in place too as we deploy our software on your infrastructure. There is then no need to expose your prediction API to the public web, everything can remain within your secure environment.

Once you’re connected, you will be able to generate in depth explanation of AI models built with text, numerical, categorical or images data immediately.

If you’d like to find out more about our XAI, please get in touch.

 

Keep in touch by joining our mailing list here.

Back to blog

Get in Touch