On Jan. 20, the start-up DeepSeek released DeepSeek-R1, an artificial intelligence (AI)-based chatbot that rivals the quality of ChatGPT and other large AI companies and products.
Hrishiraj Ramachandran is the CTO at A3Logics and founder at Keystride Digital, a software development company with 475 employees, which, among other things, provides deep tech (advanced technology) solutions. Ramachandran believes that DeepSeek gained popularity due to its higher processing speed specifically in areas such as coding and mathematical computations.
“[DeepSeek] focuses on text based tasks, whereas ChatGPT can process images, videos and voice,” Ramachandran said. “DeepSeek has been gaining popularity because it’s faster. Additionally, it offers free open-source models such as the Deep Seek-V3”
The different model that DeepSeek runs on is likely to account for these increases in accuracy and processing speed.
While ChatGPT was trained on a large amount of data from around the web, DeepSeek prioritizes “ domain specific accuracy with from subject matter experts, which enhances its accuracy,” according to Ramachandran.
Knowledge of how an AI model is trained explains how DeepSeek was able to release an AI with capabilities far beyond its competition.
Cameron Davis is a senior at Sage Creek who has completed the computer science pathway and has a special interest in AI.
When training AI, there is “an agent, which oversees the model. When the model makes positive actions, the agent reinforces this behavior typically by using a points system,” Davis explains, “Training LLMs (Large Language Models) is similar to how you might train a dog, rewarding good behavior to encourage it,” he said.
Perhaps the most influential advancement in the AI industry that DeepSeek made was that it open-sourced its AI.
“They made it open-source and free. This allows anyone with enough computing power to set up DeepSeek on their own computer,” Ramachandran said.
DeepSeek being open-source makes it easy for other AI-companies to learn how to improve their products. All of the advancements that DeepSeek made will be felt all around the artificial intelligence industry.
“The advancements they’ve made in AI will impact how I use AI in my daily life,” Davis said. “As other models incorporate these improvements, the AI I use will get better.”
Advancements in the artificial intelligence industry do raise some moral questions and dilemmas. The release of DeepSeek only intensifies these concerns.
“A concern I have is that the irresponsible use of AI could result in future generations relying less on their own intelligence,” Ramachandran said. “In the past, we didn’t have AI or machine learning tools to help us solve problems or make decisions. We had to rely on our own thinking and problem-solving skills. Now, platforms like DeepSeek are allowing people to solve problems without using their own intelligence.”
Additionally, ulterior motives might be at play. ChatGPT and DeepSeek are both for-profit businesses with shareholders to please.
“An important topic to consider when discussing AI models is how companies use user data,” Davis said. “OpenAI uses user data to train its models, but it also might be selling user data. It’s important to think about how companies like Deepseek are handling user information.”
Additionally, over two months after its release, more truths about DeepSeek have been revealed. Its $4-6 million training cost was a gross misrepresentation, with an actual cost in the billions. OpenAI still dominates the artificial intelligence landscape, with DeepSeek having nowhere near the weekly users (400 million compared to 15.9 million)
Regardless, DeepSeek has made humongous innovations in the artificial intelligence landscape and is an important step in the evolution of AI-technology.
