What Job Postings Tell Us About AI's Future

Despite the barrage of AI-related content that’s been produced over the last several months, actually understanding its impact and potential remains a challenge. It feels to me like everyone knows that generative AI represents a monumental shift in how jobs are done (hence the impulse to write and think about it) but it’s not yet clear how that shift will take place. Everyone’s taking their best guess.

One of the more perceptive reads, I think, comes from David Rotman at the MIT Technology Review, because it shows the many directions that generative AI (like ChatGPT and DALL-E) can take as its transformation moves through the economy—and the bigger changes yet to come.

Avi Goldfarb, an economist at the University of Toronto, is quoted in the piece and he encapsulates the current situation well: “[Right now, AI will] increase efficiency—it might incrementally increase productivity—but ultimately, the net benefits are going to be small,” he says. ChatGPT, of course, allows users to generate large amounts of usable text very quickly, and right now that’s often being used by workers who would otherwise be generating similar text themselves. But Goldfarb also adds that the potential is much greater: “Once we figure out what good writing at scale allows industries to do differently…that’s when we’re going to experience the big productivity boost. But if that is next week or next year or 10 years from now, I have no idea.” 

Avi was the right person to ask; he’s done extensive research on AI and I’ve had the privilege of collaborating on some of it with him. Our February 2022 paper on whether AI can become a general purpose technology is most relevant. Looking at historical trends in job postings, we were able to see that the pattern of AI jobs is similar to past patterns for jobs related to the internet, for example, which has ultimately become a fundamental presence in society and the labor market. AI is on a similar track—the time will come when it’s part of almost everything, taken for granted—though like Avi said, there’s no telling when (I wrote a bit more about that paper a few months ago, if you’re interested). 

Ultimately, the meaningful impact of AI will come not from replacing tasks that people are already doing, but by creating and completing new tasks that nobody is yet able to do. I share that perspective with my friend Erik Brynjolffson, who makes that case better than I can, and he’s been saying it for a long time—including in this article from 2017. 

But we’re not there yet. In the meantime, let’s look at what’s actually happening with AI in the labor market. Thankfully, we don’t have to speculate: there’s real data to work with. 

Searching for job postings mentioning generative AI or related skills, we can see the number is still relatively small, but growing fast. The biggest jump was between 2021 and 2022, which makes sense remembering the surge in popularity image generation tools saw last summer and ChatGPT’s similar wave in December.

With regard to specific titles and employers, it’s worth noting that the top ten for each are very focused on AI itself, as opposed to workers focused on other tasks but incorporating AI into their process. When OpenAI is hiring engineers to work on ChatGPT, they’re requiring knowledge of generative AI, but if someone is hiring a designer for their marketing department, they’re not requiring that designer know how to create images using something like Midjourney (at least not yet).

So right now it’s mainly AI-focused businesses that are explicitly laying out AI requirements in their job postings. As machine learning goes more mainstream, we’ll see more and more postings spreading to new titles and new kinds of businesses. We’ll keep an eye on the data.

Our last stop on today’s survey on the current state of artificial intelligence is the Stanford AI Index, which came out last week. The index is an annual release and it’s a big one: 386 pages covering everything AI-related, from carbon emissions to bias controversies to legislation. 

Lightcast contributed data on the labor market, which indicated that the level of AI postings increased in every industry from 2021 to 2022, except one: Agriculture, Forestry, Fishing, and Hunting saw a decline of 0.02%. 

Our findings also showed how the geography of AI hiring continues to shift. The United States has the highest concentration of AI job postings, followed by Canada and Spain. In the US, 2.05% of all job postings are AI-related. California, Texas, and New York are the US states with the most AI job postings, but Washington DC, Delaware, and Washington state have the highest ratio of AI job postings to all other job postings. 

To identify an AI job, we looked at the skills a posting required for skills related to AI, and sorted those into several clusters. ChatGPT wasn’t part of those calculations, but I’ve included it below for comparison. Demand is increasing, though it’s still very small compared to the others. I’m interested to see what the postings data continue to tell us as AI becomes more widespread; for now, though, be sure to visit the Lightcast blog for a deeper look at all our findings included in the report.

Economic News

The most important data release this week was yesterday’s CPI: inflation rose 0.1% in March, representing a 5.0% increase year-over-year. That’s the lowest yearly figure since May 2021, down from a peak of 9.1% annual inflation last June.

Combined with the encouraging numbers from JOLTS and the Employment Situation last week, where openings went down without layoffs going up and the rate of job growth cooled nicely, the overall labor market is moving in the direction we want it to go. In fact, I don’t see a need for the Federal Reserve to increase interest rates when it next meets the first week of May. We’re on track for a soft landing—which means it’s not time to change course.

Finally, some exciting news from Lightcast: we’ve made our first acquisition! Gazelle is a business-intelligence platform with software that can identify which companies are growing, based on the idea that growing companies need more of everything—and now it’s part of Lightcast. This means our clients can make the seamless transition to labor market data to strategic insight, all in one place. I’m glad to welcome my new colleagues from Gazelle aboard, and I look forward to seeing what we can do together.

Until next week, 
Bledi