Analyzing AI:Q&A with Wenjun Zhou
Wenjun Zhou, Lawson Professor of Business and PhD program director in Haslam’s Department of Business Analytics and Statistics, is currently serving as president of the INFORMS College on Artificial Intelligence. In this leadership role, Zhou is in a unique position to explain how artificial intelligence (AI) will shape the future of analytics.
AI has been in the news a lot lately, but is it really new?
In the form of intelligent systems, AI is already everywhere and has existed for a long time. A few examples include database marketing (predicting churn), recommender systems (when an online retailer suggests similar items), and search engines.
What about ChatGPT?
The arrival of OpenAI’s ChatGPT made a huge impact on many fields, including analytics. Surprisingly, it’s not very good with statistical analysis. Because it’s based on generative AI language models, it’s best for creative work such as writing and image creation. Still, it can serve as a very good personal assistant for a data researcher—generating SQL queries, explaining concepts, summarizing information, and even coming up with ideas.
How can companies use AI in a way that safeguards their (and their customers’) privacy?
Obviously, you can’t upload sensitive data to ChatGPT. To use the capability of large language models (LLMs) without disclosing internal data, companies can host an LLM in-house. They can either create an internal version of ChatGPT (for example, UT Verse) by partnering with OpenAI, or download an open-source LLM (for example, Meta’s Llama 2) for internal use or fine-tuning.
Is it still important to learn traditional statistical methods?
Yes, our students still need to understand traditional methods as they provide foundational knowledge. AI tools still have many limitations. As humans, we need to understand how our data is organized, ask the right questions, and evaluate whether the answer provided by AI is accurate. I look at AI as a support tool, not as a replacement for human analysts.
How do you think AI will continue to develop?
Generative AI is quite amazing. It will be really useful as a virtual assistant and, in the future, we might even have physical AI-based assistants to serve people and do work that is dangerous for humans. Everybody—not just analysts—must adapt to the changing environment or risk being left behind. For analysts, AI makes it possible to run queries faster, so expectations for productivity will increase.
Can AI analyze data?
Yes, especially unstructured data like text. For example, companies can use generative AI to analyze the sentiment of online reviews and summarize issues from customer service transcripts. In the past, we could build a list of words to automate this process, but the AI model is more accurate and the interface is very easy to use.