DeepSeek changes the trajectory of AI services, UB experts say

By Kevin Manne

BUFFALO, N.Y. —  The launch of DeepSeek-R1, a new artificial intelligence model from Chinese company DeepSeek, has upended the technology and financial sectors, and the University at Buffalo School of Management has experts available to discuss its implications.

To interview an expert, contact Kevin Manne, associate director of communications, at 716-645-5238 or kjmanne@buffalo.edu.

The technology

Shaojie Tang is professor of management science and systems in the UB School of Management, and is faculty director of the school’s Center for AI Business Innovation. He studies machine learning, responsible AI and combinatorial optimization. Tang says that DeepSeek is a revolutionary model because of its open-source nature and cost-effective deployment. “What makes it even more intriguing is its origin story,” he says. “The company initially aimed to leverage AI and machine learning for quantitative trading, but the development of their large language models was somewhat of a byproduct of that focus. Despite this unexpected trajectory, their models have now evolved into powerful tools in their own right.”

Dominic Sellitto is clinical assistant professor of management science and systems in the UB School of Management, and is assistant faculty director of the school’s Center for AI Business Innovation. He researches cybersecurity, digital privacy, AI usage and protection and business analytics. Sellitto says this model’s ability to “think” through problems and reduce hallucinations, combined with its free availability, has significant implications for the AI industry and subscription-based AI services. “DeepSeek-R1 can second guess itself, ‘think’ about the nature of problems, and even have an internal dialogue about how it might be wrong before ultimately arriving at an answer with an explanation,” he says. “In my estimate, this is going to drive the current AI leaders in the U.S. to quickly develop and release next-gen models.”

Kevin Cleary is clinical assistant professor of management science and systems in the UB School of Management. He brings to the classroom nearly 20 years of industry experience in information security, enterprise infrastructure and IT management in higher education and the private sector (from small startups to large publicly traded companies). Cleary can speak about how the competition from DeepSeek and its R1 model will drive innovation and challenge incumbents to address their computational and energy consumption costs. “Contemporary AI models have an insatiable appetite for specialized GPU hardware and, in turn, electricity consumption,” he says. “The long-term efficacy and qualitative advantages of this model will take some time to solidify, but this will at least challenge U.S. competitors to consider computational and power efficiency as important benchmark elements. DeepSeek will lower the hardware barriers to entry for those who wish to build their own, private, generative AI systems and take advantage of this free, open source alternative.”

Market impact

Cristian Tiu is chair and associate professor of finance in the UB School of Management. His research interests lie in empirical and theoretical investments. Specifically, he is interested in determinants of performance and asset allocation of nonstandard investors, such as hedge funds and university endowments, and how economic instability or market segmentation is reflected in the asset prices. He can speak about how with DeepSeek, an outsider has created a product that is similar — or even better — than very large players like Meta and OpenAI have, with a relatively low cost and a small team. He says that deploying these models locally is still a problem. “To make them fast, you still need the Nvidia GPUs. Using the largest and best version of DeepSeek takes a lot of computing power, and significantly more than a personal computer — up to as many as five H100 Nvidia GPUs at a cost of about $25,000 each.”

Scott Laing is clinical assistant professor of finance in the UB School of Management. He studies environmental social governance, sustainability policy, asset pricing and corporate finance. Laing can discuss the market implications of DeepSeek, and how big tech was banking on having a monopoly on the AI space, but now has overseas competition that is significantly more cost effective. “As companies have begun rapidly investing in AI, many will now have to answer investor calls on keeping spending realistic — these companies no longer have a ‘blank check’ to be innovation leaders,” he says.