IBM has launched a software service that scans AI systems as they work to detect bias and provides explanations for decisions made by AI.
The new IBM Cloud-based reliability and transparency model works with models built from a variety of machine learning frameworks and AI-including IBM's own Watson technology, as well as Tensorflow, SparkML, AWS SageMaker. and AzureML.
SaaS (fully automated) helps explain decision making and whether AI is biased - so when decisions are made - it will be possible to identify possible outcomes. not fair.
It will also automatically suggest data to add to the model to minimize any deviations that have been detected.
The explanatory power of AI decisions includes identifying the factors that influence decision-making; confidence in the proposal; and the factors behind that confidence.
IBM also said that the software keeps track of the accuracy, performance and fairness of the AI model, along with the pedigree of the AI system - meaning that it can be easily retrieved and retrieved if needed. For service or to comply with regulations and requirements.
For an example of compliance, the EU's GDPR framework relies on automatic decision making and includes the right to be heard explaining in detail how algorithms operate in a number of fields. Incidentally - that means businesses may need to check their AI.
The IBM AI scanner tool provides an automated decision analysis through dashboards - so you do not have to have IT skills to read them.
IBM is also not the first company to discover business opportunities around AI bias. A few months ago Accenture released a tool to identify and fix AI to help them be more fair.
In addition to launching the AI payment engine, IBM said its research division will launch a suite of open sourcing tools to detect and mitigate bias in AI - with the aim of encouraging process "global cooperation to solve this problem".
"IBM is currently leading the industry in establishing reliable and transparent guidelines for the development of new AI technology. We are bringing new solutions to AI businesses that face a lot of potential risks to any wrong decision. "- David Kenny, SVP of Cognitive Solutions at IBM said in a statement.
No comments:
Post a Comment