In order to push the envelope, we realized we needed a sensible way to examine AI techniques and technologies. We came up with ExCITE. Now, we’re helping you evaluate AI for your use cases. ExCITE is a technology-agnostic evaluation methodology focused on transparency in Artificial Intelligence solutions.
Learn Morewhy a prediction was made.
Why a prediction or decision was made, in terms that humans can understand.
why a prediction was made.
How long it will take to produce a prediction to better understand outcomes.
why a prediction was made.
How the system handles the data from internal processes to results.
why a prediction was made.
Predictions back from decisions and actions to specific trained data records.
why a prediction was made.
Learned records for refinement and error correction after training and deployment.
that raises the expectations of AI solutions to meet regulatory standards and encompasses many of the elements an AI needs to operate transparently and effectively in applications that affect life, death, and well-being.
The ExCITE Methodology for Verification & Validation of Safety-critical Artificial Intelligence.
When you are looking to incorporate AI you expect that it will improve your processes, reduce cognitive load for users, remove uncertainty, and evolve with your operational needs. In mission and safety-critical applications, AI has the added potential to improve safety, ensure operational and mission success, and protect the lives of citizens, ranging from such diverse areas as financial well being to physical safety. In order for artificial intelligence systems to advance both safely and effectively, they should be able to Explain, Compute, Interpret, Trace, and Edit (ExCITE) all outputs and processes.
Download White PaperOur symbolic, connectionist framework for intelligence, GAIuS™, is a meta-technology, created as a viable path and methodology toward human-level machine intelligence. The technology was designed to solve multiple problems by emulating human cognition. Because of that, the ExCITE elements are baked in as part of its design.
why a prediction was made.
Our universal input format converts user data into sequences of symbols before feeding into the system. All information flowing through GAIuS is presented in human terms so users can see what is happening and why.
why a prediction was made.
Our framework is deterministic, meaning same inputs = same outputs. Therefore, predictions are returned in a deterministic amount of time for real-time operating systems (RTOS).
how data filters through
the system.
Each algorithm and each datum can be reviewed before and after it is transformed by additional algorithms. Given that the system is deterministic, it is possible to predict specific outcomes based on adjustments in input data or parameters within the system.
predictions back to
source data.
Every learned record provided for training remains intact (i.e., not modeled) and is given a unique ID. When a prediction ensemble is returned, each prediction object contains the unique ID of the learned record the system pattern matched on to produce that prediction. A user/app can track these IDs against their application’s unique IDs to show exact training records.
source data for refinements and correction.
If a user spots a problem or something missing from the original data based on a prediction provided by GAIuS they can drill down into the sequence to make real-time, human-in-the-loop refinements. Because we do not model until prediction time, users are able to edit input data sequences without any need for retraining.