Explainable Artificial Intelligence (XAI)
For many AI systems, we don’t understand why they make their decisions. In general, as the algorithms or architectures get more complex, the decision-making processes get more opaque. This is especially true for Deep Learning.
At Aixora, we help our customers by using XAI practices to supply explanations for their AI systems .
Why Is XAI Important to My Business?
Auditing Your AI
Knowing why your AI systems make their decisions can:
Direct you toward improvements.
Help determine causes in the case of unexpected failure
Meeting Regulatory Requirements
Many US agencies, including SEC, FINRA, FTC, CFPB have new rules regarding AI.
NATO requires “AI applications will be appropriately understandable & transparent”.
The GDPR requires:
businesses using personal data for automated processing must be able to explain how the system makes decisions.
the right for users to ask for a human to review the decision of an AI system to determine if a mistake was made.
Ethical Behavior
Explanations help ensure that AI is trustworthy. Ethical companies make efforts to prevent their AI systems from causing unintended harms.
The Aixora XAI Engine
Aixora has created a software framework for explaining the outputs of machine learning models.
Our XAI engine does not require access to the internals of the models, nor does it attempt to approximate the parameters of the models. This means:
Existing systems do not have to be re-engineered to add explanations
Your proprietary knowledge remains confidential.
Reduced cost and complexity of deploying XAI
Our XAI engine can be used in the cloud and with standalone devices for IoT or non-networked use.
An XAI process illustrated. From the upper left (moving clockwise); the original teapot image sent to a CNN classifier; apply noise to find regions of most impact to the classifier; convert the regions of impact to a heatmap; overlay the original image for the result.