The rise of generative AI, like ChatGPT, Bard, Dolly, DALL-E, and many others, has made AI the topic of conversation from classrooms to boardrooms. As with any innovation, everyone has struggled to understand the potential risks and the rewards, with some people falling flat and others standing on the sidelines — I wonder which is worse.
The truth is that innovation in AI is not new, and ChatGPT is not the first or the last AI innovation we will face. More than simply using ChatGPT is required to meet the real challenge. To win and keep winning, we need to ensure we can quickly harness AI innovation but with confidence and trust in our people and outcomes.
This is where a true AI platform, like Dataiku, comes into play. It comes down to three key value propositions. First, Dataiku has always been an open and technology-agnostic platform for everything from open source algorithms to AI services and computing infrastructure. Second, Dataiku puts people at the center of AI projects. This human-in-the-loop element is crucial to success with new technologies. Finally, Dataiku starts with collaboration, oversight, and governance built in. Using the latest and greatest technology in a place where everyone is comfortable working together and has proper oversight is the secret sauce for success.
In the latest release, Dataiku 12, these traditions continue with new features that help companies confidently take advantage of AI innovation with more transparency, standardization, and centralized operations.
Increase Transparency and Explainability
One of the biggest problems for AI projects is the need for more visibility and transparency for builders, users, and executives. In Dataiku 12, new features take traditionally opaque processes and open them for everyone to see. For example, OpenAI GPT integration allows business users to incorporate Open AI's GPT models into data projects by extending datasets and performing tasks using a visual interface and natural language prompts, all while maintaining transparency and trust in project outputs.
Automatic feature generation not only saves time when developing features for modeling but does it in a visible way that gives people control and confidence in the results. With universal feature importance, data teams have a single way to evaluate feature importance across models, including those built outside Dataiku in tools like MLFlow, written in code notebooks, or created in data science tools like SageMaker. A consistent way to explain models allows for better comparison and understanding across modeling techniques.
Uplift modeling, one application of causal ML, enables data teams to focus on cause-and-effect relationships in results and for specific groups, increasing the understanding of modeling outcomes and improving results for retail and marketing use cases, fundraising, medical treatment and clinical trials, human resources, political campaigns, and more.
Universal Feature Importance Explains Models From Multiple Sources
Standardize Components and Processes
Even with the most efficient process for data science teams, building and coding everything from scratch is expensive, slow, and brittle. As advanced analytics projects move into domain teams like finance, marketing, HR, and manufacturing, having standardized components and processes becomes even more critical for success. Dataiku 12 has new capabilities to help everyone quickly onboard and reuse components and solutions.
Everyone from new data team members to business experts can benefit from the new help center in Dataiku, which centralizes a wide variety of valuable resources that provide contextual, personalized content recommendations. The updated data catalog with new data collections allows users to create curated lists of key datasets by team or use case, so everyone can easily find and share quality datasets to use in their projects. Along with version 12, Dataiku continues to deliver prebuilt, customizable industry and functionally-focused solutions that allow data and business teams to accelerate advanced analytics projects from weeks or months to days.
Help Center in Dataiku 12 Delivers Recommendations to Users in Context
Centralize Operations and Governance
As data, analytics, and AI programs expand across organizations, various teams, systems, and methods can limit trust and success. One way to manage the chaos is to centralize the operations and governance of everything from data pipelines to the most sophisticated AI projects. Dataiku 12 includes new capabilities for data and IT teams to streamline MLOps and governance processes to deploy models faster, better manage production models, and improve model governance.
A core principle of AI safety is keeping a human in the loop. Models don't always have the best or safest answer. Dataiku model overrides add a human layer of control over model outputs to enforce expected outcomes under specified conditions and guarantee compliance when results do not meet expectations. Improved deployment and monitoring streamlines and automates steps in the model deployment process, saving time and minimizing errors. New drift metrics allow teams to find problems before they impact results. New governance views, including a Kanban view, enable teams to see all projects across all stages, from development to production.
Model Overrides in Dataiku 12 Add Expert Guardrails to Predictive Outputs
The central teaching of the Greek philosopher Epictetus may be best summarized as, "It's not what happens to you, but how you react to it that matters." He taught that philosophy is not simply a theory but how you live. While things beyond your control may happen to you, how you respond to those events will decide your fate and how you will be remembered. That is something to consider as you tackle the latest and greatest AI challenges. How will you respond, and do you have the systems to ensure your success? With Dataiku 12, you should have everything you need and more.