Open AI "black box" with DeepTrace
Updated: May 5
Artificial intelligence (AI) is transforming the modern economy. PwC estimates that AI will drive global gross domestic product (GDP) gains of $15.7 trillion by 2030. As AI becomes more sophisticated, more and more industries are adopting this technology. The majority of AI algorithms are very complex. It is difficult and often impossible to understand how these algorithms are making decisions, as such these algorithms are widely known as a "black box”.
The necessity to explain these "black box” algorithms triggered the birth of a new generation of AI systems called Explainable AI (XAI). The main focus of these systems is providing an explanation of the decision-making process, so it becomes understandable to humans. The most promising areas for XAI, highlighted in the PwC report are finance, healthcare, automotive and security. Without explainability, many AI systems will not be viable commercially.
The Auromind product (called DeepTrace) provides explainable analytics for AI applications and delivers two key benefits in the decision-making process: transparency and explainability.
Transparency provides customers with insight about the nature of the data used to train the algorithm, which helps them to decide whether or not they want to deploy a model for use. For example, if the training data was biased against one gender as happened recently with Apple credit cards, then DeepTrace can identify and flag it to the customer.
Explainability provides customers with information about how exactly a particular outcome was produced which helps them to decide whether they trust the predictions. For example in the U.S., AI systems are already used for helping judges in determining sentences. If a judge doubts the reasons why an AI system recommends a particular sentence, DeepTrace can provide explanations as to how this decision has been reached increasing user confidence in the system.
DeepTrace is easy to use as a Software as a Service (SaaS) platform. It can be deployed in the cloud or on-premise if the company policy prevents integration with public cloud resources. The modular architecture of the DeepTrace solution can be supported by the biggest cloud providers and can be scaled to meet customers' demands. The availability of the DeepTrace API makes this solution easy to integrate into existing client architectures, where DeepTrace augments the current business processes rather than replacing them.
The goal of DeepTrace is to augment customer experience of using complex AI systems by providing a level of explanations necessary for building trust and confidence in human-AI interactions.