metamorworks - stock.adobe.com
Dataiku is out with a new version of its platform, updated with GPT models and aimed at making it easier for enterprises to use the power of OpenAI's generative AI models while still following responsible AI guidance.
The data science and AI vendor introduced Dataiku 12, the latest version of its platform for what it calls "Everyday AI." The platform is used to develop accessible AI applications and enables users at different technical levels to prepare data to train AI projects without code.
Dataiku 12, introduced on May 31, comes nearly a year after Dataiku 11, which featured an integration more focused on computer vision.
Updates in Dataiku 12 include new features such as an OpenAI GPT-4 integration, transparent and easy-to-understand automated feature generation, model risk project views, universal feature importance, and causal machine learning.
The OpenAI integration enables organizations to incorporate GPT-4 models into data projects with a visual interface and natural language prompts. Dataiku is not the only data science platform vendor working with OpenAI. In March, DataRobot revealed that it's working with OpenAI to incorporate Azure OpenAI into its platform.
Dataiku 12's Universal Feature Importance gives business teams ways to explain models. Causal Machine Learning helps organizations understand the reasoning behind an AI model's result. Model Risk Project Views enables enterprises to spot and mitigate risks in AI projects. Finally, Transparent Automated Feature Generation provides users with transparency and control over the AI modeling process -- the process in which AI models are created, trained and deployed.
Mike LeoneAnalyst, Enterprise Strategy Group
Dataiku's inclusion of features that provide more ways for users to understand processes in AI projects will help move those undertakings forward faster, said Mike Leone, an analyst with TechTarget's Enterprise Strategy Group.
"Dataiku is focused on ensuring AI success by accelerating time to value," Leone said. He added that this means the vendor is providing safety guardrails to ensure business users feel comfortable during AI projects.
Dataiku's new release comes as many enterprises seek ways to responsibly use generative AI tools such as OpenAI's ChatGPT product. With generative AI tools known to have "hallucinations," such as spitting out false information, enterprises are being cautious when using tools integrated with generative AI.
"There's just general skepticism toward AI and generative AI," Enterprise Strategy Group analyst Don Fluckinger said during an interview about another generative AI product last month. "'What's it going to do to our business? Can I trust it?' Some of these customers, probably most of them, are of the mindset of 'let's see what it does.'"
It is this hesitation that Dataiku wants to address. However, responsible AI is a tricky subject to tackle, Leone said.
"The biggest hurdle [to responsible AI] today is a lack of standardization," he said.
Many organizations need help figuring out where to start on a path toward responsible AI, Leone added. Organizations are also trying to narrow their responsible AI focus area to explainability, bias detection or fairness.
Dataiku is taking a multipronged approach that focuses on transparency, centralization and standardization, he continued. So they're providing those involved in the AI process with the tools to responsibly deploy and manage models.
"They're focused on enabling their customers to confidently move forward with their AI initiatives while ensuring trust and reducing risk," Leone said.
Other vendors are also pursuing responsible AI. On June 1, Domino Data Lab unveiled new features that enable enterprises to use foundation models while mitigating model risks and implementing responsible AI practices.
Esther Ajao is a news writer covering artificial intelligence software and systems.