The AI war has intensified with the latest launch of DeepSeek’s free AI-powered chatbot. It sent shockwaves across Silicon Valley and shook global tech stocks after securing the top position as the most-downloaded app in Apple’s App Store in the U.S., replacing OpenAI’s ChatGPT.
If you use generative tools like ChatGPT to improve your work efficiency and productivity, DeepSeek deserves your attention. Here’s everything we know about DeepSeek so you can use it optimally.
What is DeepSeek?
DeepSeek is a free AI-powered chatbot with features and appearance similar to the popular conversational chatbot ChatGPT. Liang Wenfeng founded DeepSeek in December 2023. While not much is known about Mr. Liang, his academic background includes an electronic information engineering and computer science degree.
DeepSeek’s R1 is a ‘reasoning’ model capable of performing coding, mathematics, and various other tasks that ChatGPT does. You can use this chatbot for content creation, learning complex concepts, problem-solving, automating tasks, etc. Users can access DeepSeek’s AI models via their official DeepSeek website or download their app from the official Apple App and Android Play stores.
How Does DeepSeek Work?
The model operates like all other conversational AI by understanding and responding to natural language inputs. Whether you need assistance with research, solving reasoning problems, drafting professional emails, etc., give it a prompt in simple language for a helpful response. Users can access any of the two Large Language Models (LLMs) offered by DeepSeek: DeepThink (R1) and DeepSeek-V3.
DeepThink (R1) is an alternative to ChatGPT’s o1 model, showing you its ‘thought process’ and the time taken to generate a response. DeepSeek-V3 works like ChatGPT’s standard model. Both are free and can be accessed via mobile apps or web browsers.
Its Standout Highlights & Impact on Consumers
The biggest differentiating factor of DeepSeek is its affordability. Unlike other top AI models, DeepSeek didn’t use massive financial resources to develop its chatbot. For instance, OpenAI’s ChatGPT incurred over $100 million in developmental costs. Conversely, DeepSeek’s R1 was made using fewer computer chips, costing as little as $6 million.
This low developmental cost has a significant impact on both consumers and geopolitics. OpenAI’s ChatGPT charges $200/month to access their premium models. The same capabilities are offered for free by DeepSeek’s R1 model. As DeepSeek is ‘open source,’ anyone can adjust and reconfigure the code according to their need. This cost advantage may force DeepSeek’s competitors to lower their prices, benefiting consumers.
The Geopolitical Advantage
The U.S. has imposed sanctions on China, restricting its imports of the most advanced computer chips. This limited supply could potentially deter China from advancing in the AI race. However, DeepSeek’s reliance on fewer computer chips can help Chinese innovators work around the imposed sanctions, lowering the sanction’s effectiveness.
Creative Training Advancements
While most advanced AI models rely on reinforcement learning to train and improve their models, DeepSeek followed a different approach. It automated the process of scoring and reinforcement learning, minimizing the need for human feedback. This process works exceptionally well for coding and math but not so well for others. .
DeepSeek used techniques similar to the ones used by Google DeepMind to train its model in 2016. All these steps outlined in DeepSeek’s open-source documentation may be followed by other companies going forward.
Here’s What Industry Leaders Had to Say
DeepSeek’s release garnered everyone’s attention in the tech circle, making it the talk of the town. While OpenAI’s CEO Sam Altman initially took a jab at DeepSeek after the model’s foundation occasionally confused itself with ChatGPT, Sam later appreciated the new R1 model. He said the model was impressive, especially for its price.
Entrepreneur Marc Andreessen stated that DeepSeek’s R1 launch is the Sputnik moment in the AI landscape. Microsoft CEO Satya Nadella also appreciated DeepSeek’s R1, calling it ‘super impressive’ at the World Economic Forum in Switzerland. Besides industry leaders, Nvidia praised DeepSeek in its official statement, citing it as an ‘excellent AI advancement.’
Prevailing Concerns Around DeepSeek
While DeepSeek’s reasoning and logical capabilities are impressive, it draws criticism for censoring topics and expressing political leanings towards a specific party while answering questions about China. For instance, when asked about Taiwan’s sovereignty, the chatbot states that the topic is beyond its current scope.
At times, DeepSeek’s R1 describes Taiwan as an ‘inalienable part of China’s territory.’ Privacy and security concerns also arise as the collected data will be subject to surveillance under Chinese law.
Countries Banning DeepSeek’s Usage
Government agencies in many countries have banned employees from using DeepSeek because of security concerns. Agencies in the U.S., Taiwan, Australia, South Korea, and Italy have blocked their government departments from using DeepSeek. Other countries may also follow suit. The decision to ban this Chinese application comes from the uncertainty of how users’ data will be stored and used by the company.
According to DeepSeek’s privacy policy, they currently collect data such as users’ email, phone number, chat history within the app, IP address, keystroke patterns, etc. This information is stored for ‘as long as necessary’ by DeepSeek and shared with service providers and advertising partners.
China’s Response to DeepSeek Popularity
While China’s Communist Party hasn’t officially released any statement, the state media is covering how the Wall Street giants and Silicon Valley are losing sleep to DeepSeek. People in China are celebrating DeepSeek’s advances as the country’s testament to self-reliance and technological prowess.
DeepSeek’s massive popularity and rise can be seen as a huge boost for the Chinese government, which seeks to build robust tech independent of the West. The company’s success has gained attention in China’s top political circles. Liang attended a closed-door symposium hosted by Chinese premier Li Qiang, which hinted that DeepSeek’s success was important to Beijing’s goal of overcoming Washington’s export controls.
