Chaos ensued on the morning of Jan. 27. Stockholders rushed to sell their shares in some of the world’s largest companies. Shares of the Nasdaq Composite, a stock index known to be notoriously “tech heavy,” plummeted nearly 600 points in response to a mass selloff of technology stocks. Nvidia had recently become the world’s most valuable company, largely due to the AI industry’s use of its chips. However, it fell 17% that day, causing CEO Jensen Huang to lose more than $20 billion—the largest single-day loss in history.
DeepSeek, a Chinese AI startup founded by Liang Wenfeng, released its long-awaited R1 model, an advanced machine-learning AI meant to compete with industry leaders. The company boasted that its new model was not only cheaper to train but also on par in terms of performance with ChatGPT’s o1. It’s reported that R1 cost only $5.5 million to train, a far cry from OpenAI’s rumored budget of $500 million. According to DataCamp.com, which ranked the AI across three benchmarks (mathematics, coding, and general knowledge), DeepSeek-R1 scored 79.8%, slightly ahead of OpenAI o1 at 79.2%. In math, R1 took the lead with an impressive 97.3%, slightly surpassing o1 at 96.4%. In coding, OpenAI o1 leads with 96.6%, while DeepSeek-R1 achieves a very competitive 96.3%, with only a minor difference. In general knowledge, DeepSeek-R1 scores 71.5%, trailing OpenAI o1-1217, which achieves 75.7%. This result highlights OpenAI o1-1217’s slight advantage in factual reasoning tasks.
However, a major drawback of DeepSeek is its heavy censorship. Like many Chinese-developed AI models, R1 has been criticized for filtering politically sensitive topics, refusing to discuss issues like Tiananmen Square, Taiwan’s independence, and government corruption. Users have reported that the model often shuts down conversations or gives vague responses when asked about controversial subjects. This raises concerns about the limitations of AI developed under government oversight, especially for students and professionals who rely on uncensored information for research. Despite this, many, including students at AHS, don’t seem to mind as long as R1 is able to outcompete and advance existing artificial intelligence programs.
“I’m in favor of more competition,” said sophomore Jordan Sheu. “China is a big player in this field and will only help these technologies go further.”
“If AI models keep improving this fast, students are going to rely on them for way more than just homework help,” said sophomore Aiden Li. “It’ll be interesting to see how schools and teachers will have to adapt.”
Teachers, however, express growing concerns about how students are engaging with AI. AP Human Geography teacher Mr. Gerry Wang warns that AI, rather than enhancing student learning, may actually hinder it.
“Students aren’t using AI to unlock their potential; rather, they’re using it to do work for them wholesale, not realizing that AI is, in fact, stunting their own growth,” Mr. Wang said.
His concerns extend beyond just students. He argues that on a global scale, nations are failing to cooperate in AI development, prioritizing competition over collaboration.
“AI has the potential to unlock human potential, but therein lies the problem: it needs to be human-driven. As long as humans are at the helm, conflict, competition, and territorialism will always be obstacles to overcome. One has to wonder why countries aren’t combining forces to use AI to cure cancer, extend human life, and solve world issues. What we’re seeing is entire states stunting their own growth in that case,” he added.
This development was especially surprising given then President Biden’s sudden announcement of a ban on the export of advanced AI computer chips to America’s adversaries on Jan. 13, just one week before the inauguration of President Donald Trump. The ban was meant to degrade China and Russia’s AI industry, but many countries, including 14 in the European Union, had their access to chips pulled as well. Speaking to reporters about the ban, U.S. Secretary of Commerce Gina Raimondo said the new rules were meant to protect advanced AI technology from foreign adversaries while still allowing partner countries to benefit from its development.
Nvidia initially appeared unfazed by DeepSeek’s success, with the company stating that DeepSeek’s R1 was “an excellent AI advancement.” This reaction seems optimistic, likely because DeepSeek still relies on AI chips—just not the most advanced ones restricted by the U.S. export ban. However, the real concern among investors is that this could be a stepping stone toward China achieving true self-sufficiency in AI development. If DeepSeek or other Chinese firms eventually find ways to train powerful models without Nvidia’s hardware, it could severely undercut Nvidia’s dominance in the AI chip market.
Investors fear that Nvidia’s rise—fueled mostly by AI chip sales—could face a reckoning if China no longer needs its technology. While DeepSeek still depends on Nvidia’s hardware for now, its ability to train a competitive AI model at a fraction of OpenAI’s cost suggests that a shift may be closer than expected.
