If AI Can Think Like a Human, How Can We Keep Up?

Artificial intelligence isn’t just evolving — it’s leaping ahead. Machines are solving complex problems, writing code and outperforming humans in areas once thought uniquely ours. At a recent Federal Reserve presentation, University of Virginia Darden School of Business Professor Anton Korinek warned: the future is arriving faster than we might be ready for.

Korinek, who last year was named to Vox Media’s “Future Perfect 50” list for his contributions to addressing the complexities of AI, said we already have AI systems that can think and reason like humans — they just can't hold as much information in their “mind” at once.

“In some sense, I would say we’ve reached AGI, subject to context window limits,” he said, referring to the amount of data the AI can refer to while generating a response.

A growing number of technologists believe AGI, or “human-level” AI will be a reality in the very near future. This begs the question: What’s our comparative advantage as humans? And how do we prepare for this uncertain future?

AI Now Outperforms Humans in Key Research Tasks

A lot has changed in just six months. In the fall of 2024, AI entered a new phase that Korinek calls “the paradigm of reasoning models.” Unlike earlier large language models (LLMs) such as ChatGPT, these systems address many of the shortcomings of the first wave of LLMs, including so-called hallucinations and difficulties with basic analytic tasks.

Now we have the rise of AI agents, or systems that automate and perform complex tasks that would normally require humans.

“They can engage in strategic planning, they can use a certain extent of long-term memory, and they employ external tools like visiting the internet and clicking stuff, or using a compiler, writing their own code, and then executing their own computer programs,” said Korinek.

These AI agents build on the increased reliability of generative AI systems and their ability to process longer texts and to operate much faster, Korinek said.

He remarked that, as an economist, he couldn’t help but analyze the comparative advantage of AI models, even amid rapid change.

“At this point, in April 2025 — and this is moving so fast — I would say that the leading AI systems have brought general world knowledge, but maybe not quite as deep as we humans in a specific context,” said Korinek. “They already have superhuman performance in processing information within their context window, meaning within everything that we can upload and process at once, and that's a pretty limited amount of information.”

And the comparative advantage that humans enjoy (for the time being)?

“We have significantly narrower world knowledge,” he explained. ”I know much less about, say, quantum physics, than the leading large language models, but our specialization means we can go and be deeper. So right now, I believe that I still have more specialized knowledge in my domain of expertise in economics. And of course, one of the big advantages of being human is that our knowledge persists, and we don't wake up with an empty, blank slate context window every morning.”

What this means, he said, is that we should “let AI do what the AI is good at, and we humans should focus on what we are good at, and that's going to make for the most productive working relationship.”

These rapid advancements have produced AI systems that can already outperform humans in some domains.

“The latest reasoning model by OpenAI can solve questions at the American Invitational Mathematics Examination (AIME) at a rate of 99.5%, which would place it essentially among the top 10 to 20 humans every year,” said Korinek.

He was candid about the impact on his own field: economic research. Tasks that once took hours or days are now being automated in minutes. He offered two examples: log-linearizing a complex economic equation and simulating a Ramsey growth model.

“I asked OpenAI’s o1 to log linearize an equation that essentially represents an arbitrage relationship between short-term and long-term interest rates and, for those of you who’ve done log linearization, it’s a pretty painful, arduous process,” said Korinek. “The AI solved it in 53 seconds and gave me the correct result. My co-author didn't want to believe it and did it manually. It took him about two and a half hours, and he got the same solution.”

Last year, Korinek used o1 to produce code to simulate the complex Ramsey growth model, but every AI system he tested, failed. Today it’s a very different story. All the reasoning models, he added, “aced” the test. “o1 Preview had to think just for 34 seconds and then correctly simulated the Ramsey growth model and gave me beautifully written and documented code.”

Still, progress comes at a cost. Solving a single FrontierMath question, Korinek noted, sometimes requires an AI to produce so much text, or tokens, that “it amounts essentially to writing several hundred books. And then at the end of it, it spits out the right response.”

What AI Means for Researchers and Students

During the Q&A session, Korinek addressed fears of college professors being replaced by AI.

If AGI arrives soon — that is, within the next two to five years — “our ability to do research is going to be devalued quite significantly,” Korinek said.

“Increasingly, I feel like whatever I can do, if I can fit it into the AI's context window, the AI can as well,” he said. “So, my only superior ability for now is that I can make connections beyond that context window, beyond, let's say, the three papers I upload into the AI.” He added he also knows data sources and has real-world context the AI can’t access.

Korinek said the advantage that a human researcher has over an AI model “is shrinking fast,” adding: “We don't know how fast it's going to continue to shrink in the future. That's the million-dollar question.”

Education will still retain a role, he added, because of its civic value — preparing students not just for work, but to become better educated citizens. Still, he said, “the fact that college graduates won't necessarily earn the same skill premium that they earn today will probably reduce demand for education and, by extension, demand for college professors.”

Students will need to adapt quickly. “In the short term, there is one winning strategy, which is whatever you are interested in, whether that is doing economics, or producing movies, or studying quantum physics, learn how to effectively use AI for doing that."

Mastering AI tools will be critical, at least "for the next year or two."

Beyond that? “All bets are off,” Korinek said.

It’s Already “Crunch Time”

Korinek acknowledged significant uncertainty about how far and fast AI will evolve. The best approach for dealing with this uncertainty, he said, is scenario planning.

He proposed three futures:

1. A business-as-usual scenario, where generative AI delivers a modest productivity boost, similar to the internet boom, but doesn’t fundamentally change the world;
2. A gradual move toward AGI over 10–20 years; or,
3. The “fastest, aggressive” scenario where AGI arrives within two to five years.

Korinek said “a lot of people believe that [scenario one] is what's going to happen,” but that may be misguided. AI could advance rapidly to the point where it can perform every task the human brain can. “The third scenario is probably the closest to the predictions coming out of the frontier labs,” he added.

Each would have different implications for economic growth, wages and labor markets.

He urged organizations — and individuals — to act by asking: If AGI is achieved within the next two to three years, what do we need to better understand now? And how can we prepare?

“If the predictions coming out of the frontier labs are right,” Korinek said, “then it's crunch time now.”

WATCH THE PRESENTATION: Professor Anton Korinek delivered a live presentation, “The Economics of Transformative AI,” to the Federal Reserve Bank of San Francisco on 22 April 2025. Afterwards, he answered live and pre-submitted questions with host moderator, Huiyu Li, co-head of the EmergingTech Economic Research Network (EERN) and research advisor at the Federal Reserve Bank of San Francisco.