The AI chatbots we invited into our long-running Axios AI+ women's bracket challenge this year were no match for most of the competitors among Axios readers and staff.

State of play: Not only did the AI chatbots have trouble filling out their March Madness brackets, they ended up finishing below most of the human pickers, at least for the women's tournament, which wrapped up Sunday.

Why it matters: AI bots are likely to become key helpers in every realm of human activity — fantasy sports included — but all that computing power doesn't guarantee better results, this year's experience shows.

Zoom in: Human entries led throughout the tournament, but ChatGPT had a chance to rise toward the higher end if South Carolina prevailed against UConn.

• With UConn's 23-point win Sunday, however, ChatGPT fell to 30th place out of the 46 completed brackets.
• Anthropic's entry, which had UCLA winning it all (they lost in the Final Four round), nonetheless beat out OpenAI, finishing in 24th place.

Another AI entrant from 4C Predictions (entered directly by the company) also had South Carolina winning it all.

• However, with better predictions in the early rounds it finished tops among the bots, landing in 18th place.

Yes, but: My bracket was even worse! I finished in 32nd place, below both ChatGPT and Anthropic.

• And some AI systems, including the ones from 4C Predictions, did quite well on the men's side.

What they're saying: 4C Predictions CEO Alan Levy noted earlier this week that, while its engine didn't get the women's Final Four exactly right, it did correctly predict the teams in the men's Final Four (all No. 1 seeds).

• "This year's results highlight a growing trend: machine learning is consistently outperforming human intuition," Levy said in a statement to Axios. "While it's true that the Final Four teams were among the favorites, the true proof of AI's superiority lies in the overall bracket accuracy, which hit 80%."
• Levy says that AI's success in the early rounds "shows that AI isn't just guessing favorites — it's identifying patterns and probabilities that human intuition tends to overlook."

Between the lines: More importantly, Levy says, AI makes its choices without getting emotional.

• For example, Levy said, his firm's AI picked Houston over heavily favored Duke in the men's tournament — and Houston beat Duke in an epic comeback Sunday.
• "This wasn't based on hype, fan sentiment, or past glory — it was a cold, data-driven decision that removed human bias," he said.

The U.S. is still the global leader in state-of-the-art artificial intelligence, but China has closed the gap considerably, according to a new report from Stanford.

Why it matters: Many leaders in Silicon Valley and D.C., including in the Trump administration, say winning this AI competition is critical to the future of U.S. national security.

Driving the news: Institutions based in the U.S. produced 40 AI models of note in 2024, compared with 15 from China and three from Europe, according to the eighth edition of Stanford's Artificial Intelligence Index, released on Monday.

• However, the report found that Chinese models have rapidly caught up in quality, noting that Chinese models reached near parity on two key benchmarks after being behind leading U.S. models by double-digit percentages a year earlier.
• Plus, it said, China is now leading the U.S. in AI publications and patents.

By the numbers: One area where the U.S. continues to dominate is in private AI investment.

• According to Stanford, the U.S. saw $109.1 billion invested last year — nearly 12 times China's $9.3 billion and 24 times the U.K.'s $4.5 billion.
• Generative AI accounted for $33.9 billion globally in private investment, up 18.7% from the prior year.
• Adoption is also on the rise, with 78% of organizations saying they were using AI in 2024, up from 55% the year before.

The big picture: The competition has narrowed not only among nations, but also among top companies, according to the report.

• As of late 2022, OpenAI and Google were in a small group of companies with a clear lead. Today, there are also credible rival models from Meta, Anthropic, xAI and others.

The bottom line: "The race is tighter than ever, and no one has a clear lead," Stanford said in its report.

Data centers, a physical manifestation of the AI race, are spreading across the U.S., with plans for massive additional investment and construction.

• Why it matters: Innovation infrastructure, once concentrated in coastal bubbles, is sprouting nationwide.

The world's densest data center hub, Data Center Alley, is in Northern Virginia.

• 40% of U.S. data center employees are in five states: California, Texas, Florida, New York and Georgia, the U.S. Census Bureau says.

• Meta on Saturday released the first in its Llama 4 family of multimodal AI models.
• OpenAI is reportedly testing a way to watermark images created by ChatGPT's new image generator. (Bleeping Computer)

Speaking of the AI+ Women's NCAA Tournament Bracket Challenge, congrats to Axios' transportation correspondent, Joann Muller, who had the top score, and to Pratham Dalal, who has the highest reader score and wins our Axios prize. (Please be in touch with us, Pratham. Sorry, Joann.)

• Also, honorable mentions to my college friend David Fischer, who had the next-highest score, and my cousin Kurt Fried, who would have won had South Carolina prevailed.
• Better luck to everyone (and everybot) else next year.