Dataiku Inc., creator of a data science platform targeted at artificial intelligence development, said today it’s adding a host of new features to its flagship offering to help customers keep up with the pace of innovation in the rapidly evolving market.

The new features are focused on generative AI and include an integration with everyone’s favorite large language model, namely OpenAI LP’s GPT-4, which powers the chatbot sensation ChatGPT.

Dataiku’s platform is used to develop AI applications and is aimed at easing related tasks, such as deploying newly created neural networks in production. It’s designed for both technical and nontechnical users, and to that end it provides tools that make it easy to prepare data for AI training without writing any code.

The tool can remove duplicate business records and make other necessary changes so it’s ready to be fed into whatever model the user is building. For experienced users, Dataiku also makes it possible to customize the data preparation workflow with custom code.

Dataiku said generative AI has become a huge priority for almost every business. However, it said, companies need to balance the urgency to deploy such models with the need to manage the potential risks of AI, which stem from poor visibility and inconsistent processes across data, analytics and the AI tools various teams use.

According to Dataiku co-founder and Chief Executive Florian Douetteau, trust via transparency and governance is therefore essential for businesses innovating with new AI models. “Dataiku 12 directly addresses this need,” he said.

The OpenAI GPT integration makes it possible to incorporate GPT-4 into various projects by extending datasets and performing tasks using a visual interface and natural language prompts, the company said.

The platform also introduces new Causal Machine Learning capabilities that help users to understand why their models generate the responses they do. Meanwhile, Universal Feature Importance helps centralize AI model explainability so users can explain their models with increased confidence.

In order to help users maintain oversight of their AI models, Dataiku has added a Model Overrides feature that enables the creation of strict rules to enforce model outcomes. To increase safety further, the platform gains a Model Risk Project Views feature that makes it easier to spot and mitigate risks associated with AI models. Finally, there’s a new Transparent Automated Feature Generation tool that provides more transparency and control over feature engineering, helping users understand where new features originate.

“We’re empowering companies to confidently explore the vast potential of AI, while providing robust mechanisms for oversight and risk management,” Douetteau said. “This fusion of power and control is the key to unlocking the full potential of AI in the most responsible way possible.”

Jed Dougherty, Dataiku’s vice president of platform strategy, appeared on theCUBE, SiliconANGLE Media’s livestreaming studio, at last December’s AWS re:Invent conference, where he talked about the company’s aim to help everyone become a data analyst:

Image: Dataiku