Ad
SAS CTO Bryan Harris during his keynote at SAS Innovate in Orlando, Florida. Image: Drew Robb/TechnologyAdvice
This year’s SAS Innovate event in Orlando, Florida showcased the evolution of artificial intelligence in the business world. Since ChatGPT unleashed generative AI (GenAI) in late 2022, business and user expectations about the benefits and potential of GenAI have often been unrealistic.
“Many people thought that all you needed to do was sprinkle a little GenAI, and all your problems would magically disappear,” said SAS CTO Bryan Harris during the keynote. “GenAI is incredible technology, but it has definite limitations. Serious issues can arise when it is not addressed to the right use cases and subjected to proper controls.”
He sketched out the evolution of technology from the PC to the cloud to general AI and GenAI. The next stages are agentic AI and advanced digital twins.
DOWNLOAD: AI Quick Glossary from TechRepublic Premium
More must-read AI coverage
Combating GenAI bias with agentic AI
An example highlighted some of the problems that can crop up with GenAI. A recent study looked into large language model (LLM) recommendations when used in a loan application setting. Regardless of the model, whether it was OpenAI, Anthropic, or Meta, bias based primarily on demographics found its way into the outputs.
“LLMs recommended far more denials and higher interest rates for Black applicants compared to white applicants,” said Harris. “LLMs alone are insufficient for most enterprise use cases. They require more robust processes, governance, and ethics.”
He cited analyst numbers that back up the fact that there is a shift going on from GenAI toward agentic AI. By 2030, agentic AI will be worth $70 billion per year, and 52% of enterprises are already implementing it, noted Harris.
GenAI alone, he said, has a human asking a question and AI providing a decision or answer. Agentic AI is based on the combination of a human-out-of-the-loop element (i.e., where AI acts on behalf of the human) and a human-in-the-loop element (i.e., where a human is engaged when needed).
“Agentic AI works on both sides,” said Harris. “AI compiles a summary for a human to review.”
He announced that the SAS Viya platform incorporates agentic AI. It offers a low-code/no-code approach to the building of agentic AI workflows for the purpose of solving complex business problems. Returning to the situation of GenAI bias in loan applications: Agentic AI within SAS Viya adds the necessary degree of governance and accountability for each decision as a way to eliminate bias. The lineage of every decision can be shown so that the correctness of the outcome can be defended. All inputs, model steps, scores, and final decisions are available to ensure accuracy.
SAS Viya graphic shown during the keynote at SAS Innovate in Orlando, Florida. Image: Drew Robb/TechnologyAdvice
Digital twins simulate AI outcomes
Harris then introduced the concept of the digital twin; essentially, it is a way to create a digital version of the physical world – this could be the workings of an engine, how an entire plant functions, or the operation of the power grid. The idea is to duplicate all areas and show it operating in real time. Once done, it becomes possible to model what-if scenarios. Instead of piloting or proving things out in a physical environment and potentially causing problems or even damage, a digital twin shows the impact of changes, innovations, or new processes before anything is implemented.
“Each business needs to define the kind of digital twin they need and its various components,” said Harris. “A digital twin enables complex systems to be seen, understood, and simulated.”
SAS is working with Epic Games to create even more realistic and workable digital twins. Epic Games has developed the Unreal Engine; as well as in video games, it is widely used in automotive, film, TV, architecture, and more to create 3D simulation environments that exactly mirror the real world. Companies like paper manufacturer Georgia-Pacific are piloting a combined SAS-Epic approach in one of its warehouses to determine the ideal number and configuration of autonomous guided vehicles (AGVs).
“Every subset of AI is fueling decision intelligence,” said Harris. “Good decisions enable you to compete and win. SAS gives you a decision advantage.”
Ad
SomaDerm, SomaDerm CBD, SomaDerm AWE (by New U Life).