What to Know about DeepSeek, the New AI Chatbot 

Has ChatGPT met its match?

Posted on February 5, 2025 in Blog Posts

Tech debates have taken center stage in American discourse in the past month. Silicon Valley leaders have inserted themselves into the Trump administration, political leaders (and their wives) have announced meme coins, a ban on TikTok briefly went into effect, Meta replaced its fact checkers, the White House announced up to $500 billion in AI infrastructure investment, and X… continues being X.

This wave of stories has been dizzying even for people who stay abreast of the news. The centrality of Big Tech to the Trump administration suggests that we’re going to see similar waves over the next four years.

Of course, not every story deserves our full attention. But one story that has been deservedly making waves is DeepSeek, the Chinese open-source AI large language model. Released on January 10, the DeepSeek chatbot, DeepSeek-R1, has created a storm of AI discourse. This post can’t fully capture the impact of DeepSeek and its criticisms as they have emerged over the past two weeks, but I hope to make the case for why this story is worth paying attention to. 

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• How Libraries Can Use Wikipedia to Combat Book Bans and Misinformation
• Librarians’ Guide to Answering Students’ Practical Questions about AI
• How One Library Cultivated Student Agency and Research in Digital Privacy

🦋 Choice is now on Bluesky. Say hi to us over there!

---

The AI landscape before DeepSeek 

As Kyle Bylin observed in his excellent post for LibTech Insights earlier this week, we have witnessed an intensification of AI discourse since the end of last year, particularly surrounding US advancements toward artificial general intelligence. Certainly, in the early days of January, OpenAI, the company behind ChatGPT, was riding high. It found a friend in President Trump, and the Trump administration’s Stargate Project would directly benefit OpenAI with its investment in AI infrastructure. OpenAI CEO Sam Altman started intoning about the imminence of the singularity with the release of ChatGPT’s newest model, o3. OpenAI was poised to hold onto its global leadership role unchallenged. 

Enter: DeepSeek 

DeepSeek threw a wrench into things. On the surface, DeepSeek looks and operates like any other AI chatbot (ChatGPT, Claude, Gemini, etc.). It has an app that’s already shot up in the app store. Its reasoning skills are said to be comparable to other current AI models. Though it has stirred geopolitical tensions with China, it doesn’t look or perform all that differently from other AI chatbot models at first blush. You can play around with it for free after making an account, but if you’ve ever used an AI chatbot before and are looking for a brand-new experience, you’ll likely be underwhelmed. 

However, beneath the surface, the creation of DeepSeek issued a powerful challenge to US tech companies, which has put the industry into turmoil: 

1. DeepSeek allegedly took less than $6 million (USD) to develop, compared to OpenAI’s $100 million for GPT-4. 

2. DeepSeek’s technical innovations mean that it uses considerably less computational power, cutting energy costs. 

3. It uses less advanced Nvidia chips, circumventing US restrictions placed on Nvidia’s AI chip sales in China. 

4. It is open source. 

While DeepSeek may not be a game-changer in terms of the outputs of its chatbot, its reduced inputs give it a huge edge in the AI arena. Indeed, despite the prevalence of “AI” as a marketing term since ChatGPT’s launch in November 2022, AI’s computational and energy costs have severely limited its application and spread. Even with premium plans, users have only a limited number of daily “tokens” to use on a company’s top AI models before they’re pushed to less advanced models. DeepSeek’s innovations in this area could lift these hard restrictions on AI’s spread. 

The fallout of DeepSeek’s launch 

Many have poked holes in DeepSeek’s official story. Experts are skeptical that the $6 million figure is accurate and suspect it does not reflect the full development costs. OpenAI alleges that, in a breach of its terms of service, DeepSeek used OpenAI’s models to train its own. If true, this would help explain the wide discrepancy in its development costs. 

Even so, it has opened a brutal question for a sector that is promised to receive major US government and private investment: Why do OpenAI and other US-based AI companies need billions of dollars in investment when a Chinese company could create a comparable product for far less? The answer to this question may likely end up being far less elegant than the question itself. However, regardless of where the facts ultimately fall on this point, their current alignment has induced a great deal of skepticism of OpenAI, which has faced a lot of criticism of its products over the years but not a comparable amount of doubt in its industry leadership. 

DeepSeek has also garnered its own controversies. Mirroring the debate over the TikTok ban, data collection and privacy concerns cast a shadow over DeepSeek. Italy has already banned DeepSeek until it learns more about DeepSeek’s data practices. Other countries, including the US, may follow suit out of national security concerns. Other critics have noticed that DeepSeek avoids sensitive topics about China, such as the status of Taiwan and the events of Tiananmen Square. Given the increasing tensions between the US and China, it’s likely that DeepSeek will quickly come to occupy a central position in geopolitics.

Further resources 

• “Is DeepSeek China’s Sputnik Moment?” (New Yorker) – Looking solely at the OpenAI–DeepSeek relationship tells only half of the story. This piece by John Cassidy positions DeepSeek within the broader US–China relationship. 

• “DeepSeek Exposed: How Good Is It Really? (Tutorial for Beginners)” (YouTube) – A side-by-side comparison of DeepSeek’s and ChatGPT’s interfaces and reasoning capabilities 

• “DeepSeek’s Chatbot Has an Important Message” (The Atlantic) – The most shocking effect of DeepSeek’s launch was the tanking of the US stock market. James Surowiecki offers a perspective on the tech business model. 

Related Posts