ETHICAL DATA SOLUTIONS
Another trend shaping the market is the rise of tools and methodologies intended to ensure greater ethical deployment of AI. “Central to the ethical use of AI is the ethical use of data,” said Jack Shu, director of sales engineering for Diveplane. “Compliance with cross jurisdictional governance can often hinder data availability. In certain cases, further processing is required, as the raw data may exhibit insidious properties such as bias or discrimination.”
Data monitoring and observability have also become important considerations, “as most data sources update continuously and may drift over time,” Shu continued. The problem faced with managing data ethically is “there are multiple technologies available to solve these challenges, although each solution tends to focus only on a sub-portion of the problem,” said Shu. “For example, the application of synthetic data allows compliance with GDPR, which enables transfer of data across jurisdictions. There are also various solutions for mitigating bias and discrimination, and yet still another set of tools for data monitoring.” This creates a need for specialized software for each of the sub-problems, “which can quickly lead to increase in cost and complexity as the business user needs to learn and maintain multiple software and models.”
Shu recommended the application of instance-based machine learning that “enables the ability to use a single platform to synthesize data, monitor the data, and, if necessary, fix undesirable properties of the data.” While such solutions are available today, “compatibility with existing enterprise infrastructure and ML investments may slow the adoption of the technology,” Shu cautioned. “If businesses can overcome the hurdles of adopting instance-based ML at enterprise scale, this would open the possibility of using a single platform for synthetic data generation, bias or discrimination mitigation, and real-time data monitoring.”
DIGITAL INTEGRATION HUBS
Today’s convoluted spaghetti mix of IT infrastructure “makes building new digital services a painstaking patchwork, with long development cycles lagging behind market demand,” said Adi Paz, CEO of GigaSpaces. A digital integration hub helps to “overcome these roadblocks by decoupling systems of record from digital applications,” he continued. “It enables rapid launch of cloud-native digital services on top of existing legacy systems, focusing development efforts on delivering a steady flow of high-performance digital services, instead of spending time on system- of-record integration.”
Such a hub is a form of “middleware” that requires “integration onto the enterprise IT architecture and connecting it to the organization’s disparate systems of record,” Paz cautioned. “Some organizations are hesitant to open up their IT infrastructure and add an external integration layer onto an existing architecture.”
Still, such hubs “facilitate business acceleration by drastically shortening time-to-market for new digital services,” and help drive innovation “by enabling developers to focus on new business logic instead of spending time on repetitive data integration tasks,” he said.