Lead image courtesy of Getty Images.
Data-Driven Business
How generative AI tools like ChatGPT could revolutionize business
The latest chatbot tech boasts exciting potential applications, as companies proceed with cautious optimism for future implementation.
Back in late 2022, few were surprised to learn that someone in Silicon Valley had finally managed to build a better chatbot. The ubiquitous pop-up technology, after all, had a reputation for being more annoying than helpful, a poor substitution for a real human agent. A chatbot that could actually raise the bar in terms of what consumers should expect? Many agreed that it was about time.
What almost no one had prepared for, on the other hand, was a chatbot of the kind eventually unveiled. ChatGPT, it soon became clear, as hundreds of thousands of people gave the tool a try, was more than just the latest iteration of artificial intelligence (AI)-powered customer-assistance software. Many used the word “revolutionary” to describe what the system was capable of doing. Others said the technology seemed sentient, like they were interacting with another human and not a machine.
Since then, the excitement has subsided, but ChatGPT has continued to impress. (Users simply type a command—“write a short song in the style of Bob Dylan“—and the bot instantly produces a response.) And now, attention is turning to the technology’s next phase: the one where businesses put it to work.
Across industries and at companies of all types and sizes, there are already countless examples of ChatGPT deployments in action.
At Amazon, for instance, software engineers have used it to write code, answer customer questions, and produce written training materials. CNET is using it to produce articles, while a company called ClearCOGS, based in Chicago, has integrated the tool into its service helping restaurants become more efficient and profitable. Ada, a “customer service automation” firm, just took the technology that underlies ChatGPT and added it to its own chatbot product. And over at the headquarters of SouthState Bank in Winter Haven, Florida: The institution’s director of capital markets says they’ve leveraged it for everything from graphic design to analyzing customer spending habits so they can make recommendations on how to save money.
In nearly every case, business leaders acknowledge that it’s still early days for ChatGPT. They’re trialing the technology, “playing with the tool,” getting a feel for what it can and cannot do. Still, many say they’ve already seen enough to know that they could be onto something big. “The world changed with the launch of ChatGPT,” noted Ada CEO Mike Murchison when the company announced it was using the technology. “The opportunity for enterprises can’t be overstated.”
From ‘wow’ to ‘wait and see’
It isn’t just business executives like Murchison who’ve tried ChatGPT and come away wondering about the possibilities.
“There’s definitely a big wow factor involved,” says Michael Chui, a partner at the McKinsey Global Institute who leads research on the impact of technological innovation on business and the economy. The technology at the heart of ChatGPT, Generative Pre-trained Transformer 3, is a so-called “large language”-processing program that relies on artificial intelligence and machine learning. Developed by San Francisco-based OpenAI and first released in 2020, the model draws on internet content “learned” during training—by one estimate, around 300 billion words, or 570 gigabytes of data—to create high-quality text in response to a written prompt.
In a report on generative AI that Chui and his colleagues at McKinsey published soon after ChatGPT was made available to the public, they point out that the technology is one of several recently released platforms that make up the “latest class” of deep learning systems. GitHub Copilot, for example, is a programming tool tailored to developers; while DALLE 2 is a text-to-image system that users can leverage to create visual content. The analysts go on to note that “business uses abound” for generative AI. In marketing and sales, operations and areas like engineering and research and development, “excitement is warranted,” they write, “but caution is required.”
Chui explains that it’s helpful to compare generative AI to autonomous driving. “You don’t want to see driverless cars on the road until you have confidence that they’ll be safe, and it’s the same with technologies like ChatGPT.” In time, could financial advisors use it to provide advice to their clients? Maybe, Chui says, but they better get it right, “because the industry has regulations, and there are very significant consequences for non-compliance.”
Challenges and opportunities
Another hurdle to consider, according to Chui, is the fact that platforms like ChatGPT are trained on data from the internet, not on information held by individual businesses. “If I want one of these generative AI systems to help me in human resources, I need it to interface with my actual HR policies and the specific benefits packages my company offers.” Large companies may have the wherewithal to do this work themselves, but most will have to pay for third-party services that provide integration through their products, he says.
Amarda Shehu, associate vice president of research at the Institute for Digital InnovAtion and a professor of computer science at George Mason University, agrees with Chui that businesses should know exactly what they’re doing if they choose to wade into the generative AI waters. ChatGPT, she says, “is state of the art” among large language models, “but you have to understand the technology behind it and be careful in how you use it.”
In her opinion, Shehu adds, most businesses would be wise to wait for ChatGPT to advance further before they attempt to use it in the real world. “I don’t think it’s ready,” she says. Consider a company that decides to use it to automate the answers it provides to customers’ questions. “If it gives false or misleading information or it prompts unethical or illegal actions, who’s on the hook for that? And more broadly, what does that do to trust and how customers see your business?”
In her lab at GMU, Shehu says, they’re developing generative AI applications to do things like predict clinical diagnoses and design new protein structures. The fundamental technology powering that research, as well as other work in “materials discovery,” is the same as that driving ChatGPT, she explains. “It’s used in everything from drug development to biotechnology and the semiconductor industry.”
Shehu says she thinks of ChatGPT “as very smart,” but far from perfect. “What OpenAI did by putting it out there was sort of a mic drop,” she says. “It was, ‘Here you go, give it a try. Go figure out what you can do with it.'”