When talking about technology, pundits often use the word “disruptive” as a compliment. In the case of DeepSeek, a “large language model” that is now challenging ChatGPT for AI supremacy, the term carries a mixed message.
Financial news about the Chinese LLM delivered a blow to tech stocks for a few reasons. First, the program appears to require less computer processing power than its competitors. Companies like Nvidia had seen their stock prices soar in recent months, because the conventional wisdom said that fueling the AI revolution would require lots and lots of specialized chips like the one the company manufactures. Such stocks tumbled when performance data about DeepSeek was published.
It is also being offered — for now at least — as a less expensive option for users. That could contribute to driving down stock prices of companies like Meta and Alphabet, who are counting on financial returns for their AI investments.
Open Source for Good
Mohamed Abazeed, a professor at Northwestern University, mostly sees DeepSeek’s disruption as a force for good — unless you’re a stockholder of a company like Nvidia. The main positive, he says, is that DeepSeek is “open source,” meaning others can play with its code and adapt it to their own specialized needs.
“It will allow developers to access, modify, and implement the model freely,” Abazeed says. “This openness is a distinguishing factor from other AI models.”
Potential Security Risk?
Daniel Khashabi, a computer science professor and AI expert at Johns Hopkins University, is less sanguine. He agrees that it grabbed headlines due to its speed and accuracy. But he is suspicious of its Chinese origins.
“While the model may appear competent based on performance in a few popular benchmarks, it may exhibit various biases and even adversarial behaviors that may not be immediately obvious to most users,” Khashabi says. “There is a clear security threat here.”
Khashabi recommends that developers who opt to work with DeepSeek "use significant caution and care." They should closely monitor its output to see that it is following their intended behavior. This is good advice not just for DeepSeek, he says, but for any software developed by a different country that has minimal to no documentation.
Article Sources
Our writers at Discovermagazine.com use peer-reviewed studies and high-quality sources for our articles, and our editors review for scientific accuracy and editorial standards. Review the sources used below for this article:
Before joining Discover Magazine, Paul Smaglik spent over 20 years as a science journalist, specializing in U.S. life science policy and global scientific career issues. He began his career in newspapers, but switched to scientific magazines. His work has appeared in publications including Science News, Science, Nature, and Scientific American.
