Answering the Hard AI Questions at IBM Think Singapore

When GenAI went public, AI stopped being a science project.

Before the launch, there were plenty of AI initiatives. Yet, many of these were driven with narrow AI objectives, often focused on department or division needs, and IT-owned. Very few looked to fundamentally change the business model, even though many acknowledged its potential to do so.

The world is different now. Consumers are warming to AI in ways never before imagined. And companies are scrambling to create use cases and scale these projects. And one company looking to steer them toward the right solutions and answers is IBM.

"In the last few years, there has been quite a fair bit of adoption in infusing AI as part of business processes. Organizations are much more open to trying out new things and extending it to the whole enterprise," says Colin Tan, general manager and technology leader at IBM Singapore.

At the upcoming IBM Think Singapore, the company is looking to lay out the foundations and framework and answer the complex business questions that companies face as AI maturity reaches a new level.

Facing the AI reality together

One of the challenges that enterprises face as interest in AI adoption rises is in costs. Without a limitless budget and during a time when investors remain cautious, it requires companies to be crystal clear about their vision of AI in their business.

At Think Singapore, Tan hopes to highlight these learnings to guide AI for business. "AI is really a journey. [It means] the vision is key, and you will need to take baby steps to get to where you want to," says Tan.

Another issue lies with the technology itself. For instance, generative AI generates new content based on past learnings. While its mistakes and inaccuracies are often brushed off in the consumer context, companies do not have the luxury. Getting accurate results can mean the difference between solid revenues and competitiveness and regulatory fines and loss of reputation.

"I think the main thing is how do you address hallucination and ensure it doesn't produce toxic results. At IBM, we address this by looking at it holistically," says Tan.

The X factor with watsonx

Tan highlights three areas that companies need to focus on with AI: How the AI manages and operationalizes processes, whether the data ingested is trusted, and what kind of governance framework you have put in place.

“That's how we came up with watsonx," says Tan, highlighting a critical solution that will headline at Think Singapore.

He explains that watsonx is based on "four core and differentiating beliefs": open, trusted, targeted, and empowering. From creating a foundation model trained on copious amounts of unlabelled data to developing a specific model for an enterprise, watsonx offers a path for companies to figure out their AI journey.

Tan notes that at Think Singapore, delegates will see first-hand the suite of watsonx products designed to get companies started on their AI journey. They include watsonx.ai, a studio for foundation models; watsonx.data to scale AI projects; and the upcoming watsonx.governance to ensure explainability—a requirement that is becoming important as regulators increase their AI scrutiny.

"We will focus on enterprise AI, which is really AI for business. We are looking to see if we can augment human intelligence and drive productivity for the enterprise," says Tan.

Such enterprise AI innovations include IBM building a foundation model for earth observations with NASA. The effort used Hugging Face to simplify geospatial model training and deployment using large-scale satellite and remote-sensing data. By publicly making this data available, IBM hopes to help companies from various industries prepare and speed up climate-related discoveries.

Recently, IBM unveiled plans to host Meta’s Llama 2-chat 70 billion parameter model in the watsonx.studio. The work continues the two company’s collaboration that already saw the release of the PyTorch machine learning framework and the Presto query engine in watsonx.data.

Getting humans to trust machines

Part of the challenge for companies in using AI is trust. There needs to be implicit trust in an AI model's recommendations, generated results, or outputs. Without this, human augmentation stalls and can, in fact, hinder productivity.

Trust is a significant area of focus at Think Singapore. Besides the watsonx.governance product, IBM and its partners will share lessons on arresting model drifts, ensuring explainability and negating hallucinations.

"AI requires machine learning; machine learning requires analytics; analytics requires the right data and information architecture. And AI will be only as good as the data that fuels it," Tan explains.

"That's why it's critical that we can identify the right data set and be able to do training based on labeled data to correct model hallucination and so forth," he continues.

While IBM focuses on trusted AI, it also creates an ecosystem with other partners to reinforce this approach.

For example, in Singapore, IBM is already working with IMDA to hire and train Singaporean professionals in emerging tech areas, including AI.

IBM also participates in the Veritas Initiative, a collaborative project led by the Monetary Authority of Singapore. The industry consortium recently released an open-source toolkit for responsible AI.

This local effort dovetails into IBM's work with the AI Verify Foundation, where the company is a premium member. “We support the foundation with our research capabilities as well as invest time with the view of contributing back to the open source community and developing AI verify testing tools,” Tan adds.

IBM Think to tackle hard AI questions

Tan encourages anyone interested in AI to attend IBM Think and use it as a starting point for their journey.

The event won’t just be about AI; it will be about AI for business, going deeper into the hard questions companies must answer today and tomorrow.

"The important thing I see working with IBM is the trust element. A lot of organizations are still coming back to IBM or working with IBM based on the trust. And similarly, for AI, we want any organization which adopts our AI or watsonx to feel very confident that they can achieve the outcome they want," says Tan.

This speaks volumes in today's market, where uncertainty and confusion reign over how AI can genuinely augment companies beyond offering small incremental boosts.

Winston Thomas is the editor-in-chief of CDOTrends and DigitalWorkforceTrends. He’s a singularity believer, a blockchain enthusiast, and believes we already live in a metaverse. You can reach him at [email protected].

Image credit: iStockphoto/cagkansayin