Is AI a Game-Changer? Yes … but What Game?

The final week of January produced a momentarily calamitous development among the investor set when the Chinese AI company DeepSeek rolled out the latest version of its chatbot. Shares of U.S. companies rebounded pretty quickly, as these things often do, suggesting that the sell-off was a bit of an overreaction.

I’m not so sure. First, some background. Coming down the backstretch of four decades in print journalism, a lot of it with newspapers, I was trained in the crucible of deadline reporting that might be labeled: “Know a Little About a Lot of Things.” More appropriately: “Pretend to Know a Lot of Things, and If Challenged, Change the Subject to Another Thing.”

It’s what I always hated about news meetings of the editorial muckety-mucks deciding what should be on the front page of the next morning’s edition. Jostling for position in the promotional track, junior editors would attempt to impress their senior overseers with their immense grasp of something they first heard about just an hour ago. The amount of ill-informed preening and posturing was always both fascinating and sickening, which is why I shed no tears when the industry ran out of lifeboats during the Great Recession.   

The thing about really good newspapering—something we don’t see much anymore—is that it broke the model by producing a well-researched, deeply insightful exploration of an issue, drawing on richly sourced data. It wasn’t just anecdotal “Outrage of the Day” reporting of this era. 

It took a long time to produce. It was expensive. And for those reasons, it’s not done very much these days.

That was the framework for my first dive into the limits of DeepSix’s grasp on things. Let’s start with something simple, says I to myself: Social Security solvency. Or insolvency. A relevant topic to about 10,000 Boomers retiring every day, and tens of millions of others about to, or already past that threshold.

So I hit Mr. Six with a prompt: “Create a model and make a projection of U.S. Social Security system solvency under a means-test maximum of $100,000 in other annual income for recipients.”

Once upon a time, exploring that premise might mean weeks at a library and waiting on snail-mailed returns from agencies responding to information requests. Then came the 1990s, and the same function was reduced to a few days of hard pre-Googling search engine work.

Mr. Six—by this point, I was sure I could start calling him by just Deep—responded with a 2,000-word assessment that provided a historical framework for the creation of a public retirement supplement program, demographic details that pretty clearly suggest an operating system akin to a pyramid scheme, assessments of the political challenges and social fault lines likely to follow from a means test, and a quick conclusion:  

“By reducing or eliminating benefits for higher-income individuals, the policy is projected to save approximately $65 billion annually, potentially extending the trust funds’ depletion by five to 10 years.”

This detailed response was returned in less than 60 seconds.

Were I assessing the long-term employment prospects with a periodical today, or considering a career as a think-tank researcher, or tasked with producing speeches for elected officials or policy papers for lawmakers, this might scare the bejabbers out of me. People who hire folks for jobs like those are going to compare the costs of a year’s salary and benefits for a policy wonk against a program that can do much of that same heavy lifting, in a fraction of the time, for little more than the cost of electricity to run your laptop query.

Are jobs going away because of this? Most certainly. But consider this: DeepSix will only be as good, as detailed, and as correct on those details as the parameters of a prompt allow.

One can still get into those aforementioned career paths, but from a different vector: The employee who understands how to get the inputs right—not just to, say, sway opinion in favor of a policy prescription, but to pre-emptively shoot down counter-arguments—stands the potential of making a very good living.

A rare combination of big-picture thinking and appreciation for detail will be required (something I hear are elements missing from the tool kits of younger workers these days). The point is, the tools of the job are evolving, as they did on the farm from horse-drawn plow to John Deere to tech-enabled precision agriculture we see today.

The end goals won’t; the path for getting there will. Employers who don’t yet understand what will be needed from an AI-reliant staff should start taking notes. Today. And furiously.