AI Snake Oil is Here, and It’s a Distraction, says Daumé

U.S. President Joe Biden, Governor of California Gavin Newsom and other officials attend a panel on Artificial Intelligence, in San Francisco, California, U.S., June 20, 2023. REUTERS/Kevin Lamarque

Whether you're on Wall Street, at your friends’ backyard barbecue, or even in the grocery store checkout line, chances are you’ve talked about AI at some point in the past few months.

The technology, which exploded in popularity with the launch of OpenAI’s generative AI chatbot ChatGPT in late 2022, is seemingly all anyone can talk about. And businesses across the economic spectrum smell the opportunity to cash in on the hype.

But the chatter about AI, and more specifically generative AI, technology that allows you to enter a prompt into a chatbot like ChatGPT and receive a human-like response, is getting ahead of itself, experts say. And that’s distracting from important discussions about the safety of the technology.

“There's a lot of snake oil out there,” University of Maryland professor of computer science Hal Daumé told Yahoo Finance. “Companies are selling how the technology works rather than what it can actually accomplish.”

The problem? With so much talk about generative AI products, discussions about its potential harms, with the exception of prognostications that the technology could lead to humanity’s downfall, risk being overlooked.

“We should be talking about short-term harms that can be foreseen…and trying to address those and then...keeping an eye on what can I do that will be useful right now,” said NYU Tandon School of Engineering professor Siddharth Garg.

AI is everywhere

I can’t seem to open an email anymore without someone pitching a “new” way of using AI. As Yahoo Finance executive editor Brian Sozzi pointed out, there’s a glut of companies pointing to AI as a means of grabbing investors’ attention.

“People who don't have the real goods will maybe have a little FOMO [fear of missing out] and try to join in the party and so that always happens,” explained Erik Brynjolfsson, senior fellow at Stanford University’s Institute for Human-center Artificial Intelligence and director of the Stanford Digital Economy Lab.

The education space is frequently targeted with products purporting to provide AI-powered capabilities.

“Companies are pretty much always touting some AI-based solution these days,” he said, adding that there’s frequently questions as to whether those solutions actually use AI and, if so, whether it’s actually helpful.

AI is such an amorphous term that companies can stretch their own definitions to cover their products. But the constant mention of new AI products means the problems that come along with the technology could be overshadowed by hype.

Generative AI is a powerful tool, but far from perfect

Generative AI’s ability to produce human-like responses to prompts makes it an impressive piece of technology for things like searching the web, putting together trip itineraries, and writing mundane emails.

But it’s that human-like element that also raises the potential for misuse. In May, an AI-generated image appearing to show an explosion near the Pentagon raised panic on social media and briefly sent stocks lower on Wall Street.

There’s also concern that generative AI could perpetuate bias across a slew of industries, whether that’s related to housing or job recruiting. When Bloomberg performed prompts asking AI-generated image platform Stable Diffusion to create photos of individuals in high-paying jobs and those in lower-paying jobs, those in the better-paying positions were depicted with far lighter skin than those who weren’t.

Still, there’s no denying that generative AI is groundbreaking. According to Brynjolfsson, the technology could lead to a doubling of productivity growth in the coming decade. Generative AI can also speed up the coding process for developers, or, as Google (GOOG, GOOGL) and Microsoft (MSFT) are hoping, reshape how we search the web.

“There's always going to be a certain amount of snake oil and people latching on to it and making unrealistic claims and false claims, etc,” Brynjolfsson said. “But I don't want people to miss that the underlying transformative impact of these technologies, the ability of them to solve a lot of problems that previously couldn't be solved is real.”

This article was published by Yahoo Finance.