Users can directly verify AI models.
Can be used in conjunction with open source machine learning models
NewNet S Neural Network Synthesis Engine to be Added Within the Year
Companies have struggled with a shortage of in-house expertise to develop and manage AI applications across various types of machine learning frameworks.

Furthermore, concerns have been raised that a more fundamental solution is needed, arguing that it is impossible to determine whether AI models can be trusted because the process by which AI derives results cannot be transparently verified.

According to a recent survey of 5,000 corporate executives worldwide conducted by the IBM Institute for Business Value, 82% of the responding companies are considering adopting AI.

However, it was found that 63% of companies are hesitating to adopt AI because they 'lack internal personnel capable of operating AI effectively.' In addition, 60% felt burdened by the adoption of AI due to concerns that "the data used for AI cannot be trusted and various internal and external regulations cannot be understood."

Accordingly, on the 18th, IBM unveiled 'AI OpenScale,' an open technology platform that allows users to directly verify whether AI models produce skewed results based on some biased data.

David Kenny, Senior Vice President of IBM Cognitive Solutions, stated, "IBM is committed to supporting customers in operating AI applications developed from various machine learning sources, while ensuring they can manage them transparently and easily." He added, "The AI economy can only be realized when enterprises trust AI and expand its applications."

With the launch of AI OpenScale, IBM expects to accelerate corporate AI adoption by eliminating the black-box-like uncertainties inherent in AI. Furthermore, regardless of the systems or operating environments where AI models are applied, companies can accurately understand and transparently manage decision-making processes derived from AI, as well as detect and rapidly respond to AI bias.
AI OpenScale makes the decision-making process easy for users to understand. It visually illustrates the process of deriving AI insights and explains it in easy-to-understand terms. Furthermore, based on powerful and automated bias removal technology, it continuously monitors AI applications to detect and resolve biases. In addition, it preserves all records regarding predictions, models, training data, and metrics to enable companies to comply with laws and regulations such as GDPR.

In addition, AI management and scalability are easy. It can be used in conjunction with open-source machine learning or deep learning models such as TensorFlow, Scikit-learn, Keras, and SparkML. It can also handle applications and models trained and hosted on IBM Watson, IBM PowerAI, Seldon, AWS SageMaker, AzureML, and non-IBM engines.

IBM plans to offer 'NeuNetS,' a neural network synthesis engine that automatically builds complex deep neural networks used in AI design, within AI OpenScale by the end of the year. NeuNetS, an AI that builds AI, is a core technology for automating AI development. Companies can automatically design and train customized AI models for specific datasets in less time than data scientists do. Ultimately, this enables companies to rapidly adopt and easily operate AI without relying on their own proprietary technology.

IBM plans to offer AI OpenScale through IBM Cloud and IBM Cloud Private later this year.