Datarobot updates AI cloud platform with enhanced bias mitigation
As global enterprises continue to adopt artificial intelligence (AI) for improving business outcomes, companies helping with developing and deploying such solutions are doubling down on their offerings to stay ahead of the curve.
Just a few months ago, H2O AI launched Hydrogen Torch deep learning training engine, and now, DataRobot has announced an update for its AI cloud platform, focusing on enhanced bias detection, code experimentation and expanded MLops for enterprises.
Enhanced bias mitigation
Bias has been a long-standing concern in AI development, particularly in areas like hiring, criminal justice, and healthcare. Yes, companies do put effort into enhanced governance and monitoring, but even today, there can be instances where models can embed human and societal biases and deploy them at scale.
To tackle this challenge, DataRobot introduced bias and fairness tools in 2020. The capabilities, among many other things, allowed companies to select data attributes they wanted to protect from bias and provided automated insights on disparity (if any). This enabled users to make necessary adjustments in their models wherever and whenever required.
At the ongoing AIX conference, DataRobot announced the plan to take these bias mitigation capabilities to the next level. The company said its AI cloud platform will now deliver a higher level of prevention by automatically identifying biases and adapting models. This way, the problem of bias would be addressed quickly and well before a broken model moves toward the deployment stage.
In addition to bias prevention, the AI development and deployment company announced fully integrated Code First Notebooks and location-capable predictive AI apps to jumpstart AI initiatives for maximum business value. The former stems from DataRobot’s acquisition of Zepl and provides data scientists with a purpose-built environment – complete with access to all necessary tools and resources – for exploratory, code-centric work. Meanwhile, the latter, driven by the integration of geospatial data, enables business users to deliver insights by location and markets for more targeted business decisions.
DataRobot is also expanding MLops capabilities of the AI cloud platform to support enterprise operations for full model lifecycles. As part of this, the company said it will create an open system and integrate it with platforms like GitHub, Sumo Logic, Splunk, Datadog and Zendesk.
Further, the AI platform will get management agents to support complex remote model deployments as well as automated compliance documentation capabilities for all models, including those built outside of DataRobot.
“AI solutions must adapt to deliver tangible business results and gain the trust of the enterprise. Today, we are living up to that promise with enterprise-ready features that expand the impact of AI across businesses, teams, industries and regions,” Nenshad Bardoliwalla, chief product officer at DataRobot, said
With companies like DataRobot and H2O AI making strides with their development tools, AI is expected to see mass adoption across enterprises worldwide. According to a recent Omdia study, 25% of enterprise buyers are looking to scale AI projects across multiple business units or functions. Meanwhile, Forrester envisions that 100% of enterprises will have AI in use (in some way) in just three years.
“Building on this kind of momentum, however, will demand that practitioners look beyond the operationalization of basic ML workflows to also view those workflows as an integral part of the business itself, fully embracing business users, speeding time to value, and minimizing business risk through tools supporting transparency, trust, and compliance,” Bradley Shimmin, chief analyst for AI platforms, analytics, and data management at Omdia, said.