IBM’s watsonx Day 2 Focuses on AI Governance
In 2023, AI went mainstream—or at the very least, techies were no longer the only ones talking about it. ChatGPT Version 4 became available in March of last year. With that, and with the ready accessibility of AI bots from Google, Microsoft and others, more people tinkered with these tools and shared their experiences. Ultimately, the public found that the technology, while intriguing, remains a work in progress.
However, another realization that’s emerged over the past year is that AI is finding its place in the business world. There are substantial challenges of course, but also, at this very moment, there are significant opportunities. This is the message that IBM delivered during its watsonx Day 2 virtual event in December.
IBM’s watsonx Day 2
“2023 was a year of opportunity for generative AI and the whole AI landscape, because generative AI proved to us that you can take large amounts of unstructured data and create value out of it,” Dinesh Nirmal, senior vice president, products, IBM Software, said during the keynote. He added, “[Now that] enterprises and individuals have gotten a taste of generative AI, 2024 and beyond is about ‘how do I adopt this in my enterprise?'”
Over the closing weeks of 2023, IBM remained active on the AI front, widely publicizing its plans and strategy—and in the process impressing one observer who knows the company well. In December, IBM reached an agreement with Software AG, acquiring a solution called Stream Sets as part of the transaction. Stream Sets will be used to provide data ingestion capabilities to the watsonx platform.
watsonx.governance
Meanwhile, watsonx itself continued its growth and evolution, as IBM unveiled watsonx.governance, touting it as a way to build “trusted and responsible AI.” During the watsonx Day 2 event, IBM’s Heather Gentile, director of the .governance product, provided an overview of the solution, which is designed to allow businesses to deploy and manage both large language (LLM) and machine learning (ML) models.
Gentile started her presentation with some eye-opening numbers, stating that “80% of enterprises are working with or planning to leverage foundation models and adopt generative AI in support of their business strategies” and that “generative AI could raise global GDP by 7% in 10 years.”
“While we’ve had AI governance technology in the market for quite a while, generative AI is definitely creating an inflection point… The speed, scope and scale of generative AI is unprecedented,” she added.
The keynote also featured a conversation between Nirmal and Gary Kotovets, chief analytics officer for Dun & Bradstreet. They discussed the long-time IBM client’s use of watsonx and how D&B manages the compliance and ethical issues posed by AI. Kotovets pointed to four “guiding principles” that must be followed when implementing AI solutions:
1. Establish a framework for compliance based on an understanding of the regulatory, ethical and legal implications of conducting AI-based data modeling.
2. Security. Establish controlled security standards and protocols that are specific to the AI environment.
3. Educate teams on generative AI, models, frameworks and organizational protocols.
4. Create a detailed roadmap with clearly defined ROI.
“Step back and survey your processes and applications and try to identify where generative AI can bring the greatest efficiency…of course combined with security and legal frameworks in place,” he explained. “Pick the right business units and the right data sets to work on in your first cases, where you’re not putting yourself too much at risk of exposing and leaking that data.”
More watsonx Days Ahead
In introducing the watsonx Day 2 event, Nicholas Renotte, chief AI engineer, watsonx team, said that the initial watsonx event held in June 2023 attracted some 5,800 viewers worldwide. He added that IBM plans to make watsonx events an ongoing occurrence, with a goal of hosting virtual conferences quarterly. Replays from watsonx Day 2 will remain online through February and can be viewed here. IBM registration is required.