DeepSeek Surges on App Store: What Is the Chinese AI Model and Why Is It a Big Deal?
China's artificial intelligence company jumped past OpenAI's ChatGPT on the Apple App Store after releasing two new models The post DeepSeek Surges on App Store: What Is the Chinese AI Model and Why Is It a Big Deal? appeared first on TheWrap.
DeepSeek sent shock waves through the artificial intelligence world this past weekend and Wall Street on Monday. If you follow the tech industry even sparingly — or happened to jump on X or the App Store recently — you likely saw the name.
But what is DeepSeek, and what is the big deal all about? Let’s jump into it.
For starters, DeepSeek is a Chinese AI company and research lab that creates open-source large language models, or LLMs. Open-source LLMs are AI models that allow others to build on its technology.
X’s Grok and Meta’s Llama are some other well-known open-source LLMs, while OpenAI’s ChatGPT is the most popular closed-source LLM.
Why Is It In the News?
DeepSeek released its R1 and R1-Zero reasoning models on Jan. 20, and their capabilities quickly gained notice in the tech and AI worlds. In particular, there are two key aspects to DeepSeek’s latest releases:
1) Refined Reinforcement
DeepSeek’s model uses a technique called Reinforcement — where a model makes decisions to achieve a certain goal — that goes beyond other models. That is because it does not use human feedback to refine itself, which other models do.
As Ben Thompson of the tech-focused Stratechery blog put it succinctly: “LLMs to date, however, have relied on reinforcement learning with human feedback; humans are in the loop to help guide the model, navigate difficult choices where rewards aren’t obvious, etc… R1-Zero, however, drops the HF part — it’s just reinforcement learning.”
2) Distillation
Another important factor here is DeepSeek’s distillation technique. Distillation is where a smaller model like DeepSeek is able to lift the information from larger models, while also using less memory and computing power.
This also helps cut down on costs. DeepSeek said its recent breakthroughs cost less than $6 million and took less than two months to build.
Why Is This a Big Deal?
From a macro standpoint, it shows that China — remember, China’s communist government is closely linked to all of its companies, especially the major tech firms that branch out into different markets — is further along in AI innovation than many had thought.
And DeepSeek’s performance is particularly noteworthy, given its low cost and its ability to navigate around chip restrictions that were rolled out by former President Joe Biden.
Marc Andreessen, co-founder of the tech-focused Andreessen Horowitz venture capital firm, called it AI’s “sputnik moment” on Sunday.
Competitors
DeepSeek has crashed an already-crowded AI space. It leapfrogged ChatGPT in the Apple App Store as the most-downloaded free app this weekend, indicating there is strong user interest in what DeepSeek can do.
Many analysts believe DeepSeek, and what is means about Chinese AI capabilities, was the main reason the U.S. stock market was down on Monday morning, with the S&P 500 dropping 1.72% in the first 90 minutes of trading.
Meta, which announced on Friday it would be investing $60 billion-$65 billion to fuel its AI business, was not hit by the Monday drop, with its share price climbing about 1.5% in early trading. Google, which has its own AI model, Gemini, did see its stock price drop about 2.5%, while Tesla was down 1%, although its business is obviously more than just its AI component. Microsoft, OpenAI’s partner, saw its stock price drop 3.7%.
The biggest hit was to Nvidia, which supplies chips and other important tech that undergirds AI development. Its stock price was down 14.8% in the first hour and a half of trading on Monday.
Censorship?
To reinforce that DeepSeek is ultimately beholden to China’s government, many noted this weekend that its model will not answer certain questions on topics the CCP is not a fan of — like the 1989 Tiananmen Square massacre. This is similar to how TheWrap recently found “Little Red Book,” aka “RedNote,” would censor content about Tiananmen Square or China’s treatment of the Uyghurs.
The post DeepSeek Surges on App Store: What Is the Chinese AI Model and Why Is It a Big Deal? appeared first on TheWrap.