Despite the hype, most jobs cannot be cost-effectively replaced by artificial intelligence, according to a study conducted by the Massachusetts Institute of Technology.

Researchers at MIT looked at how expensive it would be to automate various tasks that fuel the U.S. economy. They focused on jobs where computer vision (using computers to identify and interpret objects and people) was leveraged in industries like teaching and property appraisal, as reported by Bloomberg. The study concluded that less than a quarter (23%) of workers could be replaced cost-effectively.

In other words, the researchers concluded that replacing more than three-quarters of the U.S. workforce with AI remains too costly.

To conduct the study, online surveys were leveraged to collect information on roughly 1,000 visually assisted tasks spanning 800 occupations. The researchers concluded that 3% of these jobs could be automated economically today. However, they said this figure could rise to 40% by the decade’s end if data costs shrink and accuracy improves.

The Dallas Express recently reported that Dallas was ranked as one of the top markets most resilient to the risk of AI overtaking the workplace. Dallas’ number two ranking behind Phoenix was attributed to the city’s diverse job landscape. In contrast, smaller cities seem to be at a higher risk, with jobs typically concentrated in fewer sectors.

While AI may have a long way to go before it displaces a broader spectrum of jobs, it is not unheard of. Earlier this month, AI played a part in layoffs at Google. Parent company Alphabet announced hundreds of employees would be cut as it transitions its focus to artificial intelligence to help control costs.