The last few days have been exciting for anybody interested in technology. Tech stocks, especially those of companies dealing in artificial intelligence, tumbled. Giant companies like Meta and Nvidia faced a barrage of questions about their future. And companies like Apple questioned the logic behind overspending on AI. All thanks to DeepSeek, a little-known Chinese artificial intelligence start-up.
The accomplishment, which comes with a powerful AI model, attracted far less money than many AI experts thought was necessary. DeepSeek is “AI’s Sputnik moment,” Marc Andreessen, a tech venture capitalist, said on social media.
How to use DeepSeek?
You can access the bot on a smartphone or tablet. Simply download the app from the Apple App Store or Google Play Store or simply type ‘chat.deepseek.com’ in the browser of your choice. The tricky part has to do with creating an account, without which you can’t proceed. It took us a day to create an account because of the heavy load on DeepSeek servers.
After you sign-in, you will be created by a ChatGPT-like user interface with a text box at the bottom of the screen. This is where you key in queries. If you want to check out previous chats, tap on the two horizontal lines that appear on the top left to bring up past conversations with the AI chatbot.
What data is DeepSeek collecting?
Going by DeepSeek’s privacy policy, quite a bit of data is being collected, including chat and search query history. Also collected, keystroke patterns and IP addresses from the device the user is on. But DeepSeek is not the only one. Many other AI services collect similar data. The collection of data often leads to questions from regulators.
“Data security concerns are always a critical issue when using AI chatbots, and this is not unique to DeepSeek,” Angela Zhang, a law professor at the University of Southern California, told NPR. “Even US-based AI firms like OpenAI have faced significant scrutiny and investigations in the EU over potential data privacy violations,” she said, referring to the European Union’s strict data protection laws.
Where is the collected data kept?
At the moment, DeepSeek sends the data it collects to servers in China. In India, Union minister Ashwini Vaishnaw recently addressed the issue at an industry conference: “Data privacy issues regarding DeepSeek can be addressed by hosting open source models on Indian servers.” He said DeepSeek’s models are likely to be hosted on India’s new AI Compute Facility that is scheduled to begin operations “in the coming days”. The facility is powered by 18,693 graphics processing units (GPUs) — almost 13,000 of those are Nvidia H100 GPUs, and about 1,500 are Nvidia H200 GPUs. The Defense Information Systems Agency, which is responsible for the Pentagon’s IT networks, moved to block access to the startup’s website last week.
Are queries censored?
Since DeepSeek's chatbot surged in popularity, researchers have documented how its answers reflect China’s view of the world
Yes, queries that the Chinese government doesn’t want to address are not shown by the chatbot. For example, anything to do with the 1989 Tiananmen Square massacre will be met with “Sorry, that’s beyond my current scope. Let’s talk about something else.” We tried asking questions about the issue and even Taiwan but the queries were bypassed. The limits are a reminder that the Internet in China is government controlled and you can’t trust the answers.
Is OpenAI happy about DeepSeek?
OpenAI and Microsoft are investigating whether DeepSeek has used OpenAI’s API to integrate OpenAI’s AI models into DeepSeek’s own models, according to Bloomberg. Further, OpenAI told the Financial Times that it found evidence linking DeepSeek to the use of distillation — a technique developers use to train AI models by extracting data from larger, more capable ones. This is an efficient way to train smaller models at a very low cost. While developers can use OpenAI’s API to integrate its AI with their own applications, distilling the outputs to build rival models is a violation of OpenAI’s terms of service.
At the same time, OpenAI CEO Sam Altman is waking up to the idea that his company should consider giving away its AI models. DeepSeek’s AI models are open-source or anyone can use it. In an AMA or “ask me anything” session on Reddit, a participant asked Altman if the ChatGPT maker would consider releasing some of the technology within its AI models. Altman said: “I personally think we have been on the wrong side of history here and need to figure out a different open source strategy…. Not everyone at OpenAI shares this view, and it’s also not our current highest priority.”
Who founded DeepSeek?
Liang Wenfeng is the founder of DeepSeek. He first made his mark in China’s investment world in the late 2010s, cofounding a hedge fund that used artificial-intelligence models. In 2023, he poured money into AI and assembled a team to build China’s answer to the Silicon Valley front-runner OpenAI. His 2010 thesis at Zhejiang University took on a topic of interest: Improving intelligent tracking algorithms for surveillance cameras.
“Providing Cloud services is not our main goal. Our aim is still to achieve AGI,” he told 36Kr, a Chinese tech outlet.
His hiring philosophy is not ordinary — DeepSeek’s engineering teams have been joined by literature buffs to help refine the company’s AI models. “Everyone has their own unique journey and brings their own ideas with them, so there’s no need to push them,” he told 36Kr.
What do bigwigs think of DeepSeek?
One of the companies that has been cautious about spending on AI is Apple. The company’s CEO Tim Cook said DeepSeek’s AI models represent “innovation that drives efficiency” during an earnings call last week. “In general, I think innovation that drives efficiency is a good thing. And, you know, that’s what you see in that model,” said Cook, responding to an analyst’s question about how DeepSeek’s AI models would impact Apple’s margins. He also said Apple has always taken a “prudent and deliberate” approach to these types of expenditures.
Even Microsoft CEO Satya Nadella spoke about DeepSeek during an earnings call. “What’s happening with AI is no different than what was happening with the regular compute cycle,” Nadella said. “It’s always about bending the curve and then putting more points up the curve…. I think DeepSeek has had some real innovation.” He also said that the efforts of DeepSeek will eventually be broadly used throughout AI platforms. He compared the switch to data storage from server to Cloud. “And so when token prices fall, inference computing prices fall, that means people can consume more and there will be more apps written,” Nadella said.
Meta’s Yann LeCun, the company’s chief AI scientist, said DeepSeek’s breakthrough is good news for the social media giant. “To people who see the performance of DeepSeek and think: ‘China is surpassing the US in AI.’ You are reading this wronug. The correct reading is: ‘Open source models are surpassing proprietary ones,’” Yann LeCun said in a LinkedIn post.
Nvidia, whose stock is affected, called DeepSeek’s R1 model “an excellent AI achievement”. “DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling. […] DeepSeek’s work illustrates how new models can be created using that technique, leveraging widely-available models and compute that is fully export control compliant,” Nvidia said in a statement.
Anthropic’s Dario Amodei, who believes Chinese AI companies have more GPUs than expected, said in an interview with CNBC: “It’s been reported, at least, that they [DeepSeek] have a cluster of 50,000…. I think if the United States can’t lead in this technology, we’re going to be in a very bad place geopolitically,” he added.