AI Is Coming for White-Collar Jobs, Too

(rafapress/Shutterstock)

Think AI is just coming for customer service jobs? Think again, say AI experts, who point to recent advances in large language models as evidence that white-collar and professional jobs will be disrupted too. Figuring out how AI and humans will coexist in the workplace is shaping up to be a key conversation for 2023 and beyond.

“I think there are traditional white-collar businesses, white-collar professions that are going to be transformed by some of the innovation in large language models and AI technologies,” said Peter Wang, the CEO of Anaconda, a provider of data science tools. “And that is going to create really interesting social and cultural dynamics that will basically settle out over the rest of this decade and reverberate into the 2030s.”

Large language models, such as GPT-3 and BERT, have made inroads in conversational AI. Companies have replaced or augmented their human call center workers with AI that can understand typed or spoken requests and respond with (hopefully) helpful information. This type of technology has the capability to displace information workers too, Wang said.

“I would say to anyone who has essentially a desk job and they go day in, day out and they kind of generally do the same thing and they don’t have to think too much about it–you better watch out because that job is probably going to be automated away,” Wang told Datanami in mid-November.

“The only reason it hasn’t been is because it hasn’t been worth somebody coming along with an enterprise software development team to directly tie in precisely to the business data system that you’re clicking on over here, the data set that you’re moving into some other data system over there,” he continued. “But in the next five years, a whole pile of stuff is going to come on down, and it’s going to wire these directly to each other and then there’s no reason for you to be clocking in every day.”

(Rawpixel.com/Shutterstock)

ChatGPT is the latest AI tool to catch people’s attention. Released to the public by OpenAI on November 30 (before the interview with Wang), the new interface to the GPT-3 model quickly garnered over a million users. The service has demonstrated a remarkable ability to generate detailed responses to questions. In addition to generating written content, it’s also shown the capability to code.

Jonas Kubilius, an AI researcher and CEO at AI company Three Thirds, sees the evolution of AI models like Stable Diffusion, GPT-3, and GitHub Copilot into multi-modal models that can handle text, images, audio, and other inputs for multiple tasks. Eventually, the content emanating from these models will be harnessed into a business model that drives profits for developers and content creators, he said.

“We will start seeing a shift from using AI for static tasks like classification to language-model-driven interactive workflows that help people perform their tasks more efficiently,” he said.

However, there are also potential nefarious uses for these models. Security researchers are warning that ChatGPT’s capability to string words and script code together could also make it suitable for use as a hacking tool. Check Point Research today published a report detailing how OpenAI’s latest creation can be used by cyber criminals to execute spear-phishing attacks. (Cybercriminals, alas, are white-collar information workers too, even if they wear black hats.)

“The expanding roles of LLM and AI in the cyber world is full of opportunity, but also comes with risks,” Check Point Research wrote. “Complicated attack processes can also be automated as well, using the LLMs APIs to generate other malicious artifacts.”

AI developers typically use APIs to access pretrained models, such as GPT3. However, much of the technology behind large language models is open source, allowing developers to use it themselves. The combination of more data, open AI tooling, and better education is combining to lower the barrier to utilizing AI, according to Sri Ambati, CEO and founder of H2O.ai, a provider of data science tools.

“That’s really the biggest barrier for our customers. Not all of them have top Kaggle data scientists continuously learning,” Ambati said. “Everything to build those models is no longer limited to having great data scientists understand all the deep learning frameworks that are coming at us every day. So you can start building those models in a very simple low-code, no-code way.”

 (thinkhubstudio/Shutterstock)

Hayley Sutherland, who tracks the market for conversational AI tools and technologies as a research manager for IDC, said many people are surprised at how good conversational AI systems have gotten.

“As technology matures…you have some people who are just realizing, wow, this is smarter than I thought it was,” Sutherland said.  “Over the last couple of years organizations have really seen the return on investment for conversational AI…”

What companies are now struggling with is figuring out where AI works best and where humans work best. It won’t be feasible to replace all humans with AI, Sutherland says. In fact, that could introduce new problems to the equation.

“The best way to leverage AI may be an understanding how can we use it to augment human workers,” she said. “I think that is a balance that we’re really seeing organizations start to come through with, especially in the last year as they wrestle with what’s being called the Great Resignation.”

With a recession looming in 2023, the labor participation rate still down due to economic disruptions related to Covid, and upward pressure on salaries due to inflation, companies will have more incentive than ever to push the boundaries on what AI can do in the enterprise. Anaconda’s Wang expects companies to explore those boundaries to create new business opportunities.

“I think that we are going to see a shockingly high number of use cases that are not precise, that were not easy to automate in the past, suddenly become much easier to automate,” Wang said. “So it’s almost like a fuzzy-matching kind of thing. A lot more use cases in the long tail of use cases will be handled by some of these AI systems.”

Wang is less bullish on AI’s ability to displace creative workers, although he does see potential for AI to help creative workers by automating the dull, repetitive parts of their jobs. That could help to spur the generation of more creative content, which engages audiences in new and interesting ways, he said.

“It’s a weird world we’re going into,” Wang said. “And it’s possible that a lot of work gets displaced, and then we have to have a formal conversation about UBI [universal basic income] and productivity and what does a world look like where not only are blue-collar work and labor work being replaced by smart robots, but then a lot of traditional white-collar workers being displaced by AI systems.”

Related Items:

The Need to Refocus AI Priorities

Conversational AI Poised to Be Major Disrupter

Google Debuts LaMDA 2 Conversational AI System and AI Test Kitchen