No one would have thought that a humble side project of a Chinese hedge fund would shake the entire tech community. The rise of DeepSeek AI has caused shockwaves in the AI world, even affecting the stock market. What is more surprising is that DeepSeek AI was developed using only a few million dollarsβa small fraction of the investment made by OpenAI with $100 million and Google with $200 million.
Fuel to the fire, DeepSeek AI was released on the very day of Donald Trump’s US Presidential inauguration, wondering why. So, what is DeepSeek AI, anyway? How does it relate to ChatGPT and Google Gemini? And is it really an evolutionary AI model, or another calculated Chinese move?
Let’s try to find out in this blog.
What Is DeepSeek AI?
DeepSeek AI is not developed by a tech giant such as OpenAI or Google. Rather, it was begun as a side venture by a Chinese hedge fund manager named Feng Wen. The fund was largely involved in quantitative stock market trading, small crypto trading, and mining. They came to an understanding that their old GPUs just lay idle during weekends when the stock market was not open.
To make the most of these resources, the hedge fund embarked on a little experiment with AI model training. The tiny side project soon developed into a major development that stirred the whole AI sector. And thus, DeepSeek AI was created.
Unlike other models of AI, DeepSeek AI was not originally revolutionary. Its initial model, R10, was created entirely through reinforcement learning without supervised learning, so it was weaker than ChatGPT or Google’s Gemini. That all changed when DeepSeek R1 was launched on January 25, the same day that Donald Trump took office as president. This new model combined reinforcement learning and supervised learning, which made it considerably more powerful.
DeepSeek AI Learning Mechanism: Reinforcement and Supervised Learning
DeepSeek AI R1 is learned on a mixture of reinforcement learning and supervised learning.
Supervised Learning: Entails inputting labeled data into the AI, letting it know what is correct and wrong. It is through this that models such as ChatGPT and Google Gemini learn.
Reinforcement Learning: Takes place via trial and error, where the AI learns by being rewarded or punished.
For instance, if a kid touches an electric plug and receives an electric shock, he or she learns to never do that again. However, if he or she consumes chocolates and loves the flavor, then he or she is likely to consume chocolates again. This trial-and-error learning process is the basic principle of reinforcement learning.
DeepSeek AI R1 is special as it integrates both these methods, thus proving to be super-efficient and economical.
What Sets DeepSeek AI Apart
Affordable Training for AI: OpenAI invested $100 million and Google invested $200 million on training their AI models, whereas DeepSeek AI was trained for a few million dollars.
Open Source Technology: DeepSeek AI is fully open-source technology, unlike ChatGPT and Google Gemini, which are closed-source. It can be accessed, edited, and a new model developed from its technology by anyone.
Inexpensive AI Computing: DeepSeek AI provides considerably lower costs for AI operations than its competitors. While OpenAI is charging $1 for 1 million token inputs, DeepSeek AI does the same task for $0.55 only.
High-end Hardware Utilization: DeepSeek AI utilized older GPUs that otherwise went unused, maximizing available resources for AI training.
DeepSeek AI vs. ChatGPT vs. Google Gemini
Feature DeepSeek AI ChatGPT Google Gemini
Training Cost Few million dollars $100 million $200 million
Learning Approach Reinforcement + Supervised Supervised Supervised
Open Source Yes No No
Price for 1M Tokens\t$0.55\t$1.00\t$1.20
Launch Impact\tStock Market Crash\tMarket Stability\tMarket Stability
The affordability, open-sourcing, and novel training methods of DeepSeek AI position it as a serious player in the AI industry.
Is DeepSeek AI a Game-Changer or a Chinese Conspiracy?
One of the most controversial aspects of DeepSeek AIβs launch is its timing. Released on the same day as Donald Trumpβs inauguration, it seemed like a calculated move by China to make a statement. Additionally, some experts believe that DeepSeek AI was developed using advanced chips obtained before the US imposed sanctions on China.
Despite these concerns, one undeniable fact remainsβDeepSeek AI is disrupting the AI industry. With the US tech market experiencing significant losses following its launch, itβs clear that this AI model is a force to be reckoned with.
US Stock Market Crash After DeepSeek AIβs Launch
After DeepSeek AI R1 was released, US technology stocks crashed. NVIDIA, one of the largest firms in the AI hardware sector, saw a significant decline in stock value. Investors, who had previously thought big tech firms like Google and OpenAI would take over AI, were doubting if a tiny Chinese startup would change the overall scenario.
This change in market perception has resulted in an AI arms race, with both US and Chinese firms speeding up their AI development.
Final Thoughts: What’s Next for DeepSeek AI?
DeepSeek AI has shown that groundbreaking AI models do not always need huge funding. By utilizing open-source technology, reinforcement learning, and affordable training techniques, it has emerged as a credible alternative to ChatGPT and Google Gemini.
But the million-dollar question isβcan it keep its growth and influence? Will it keep shaking up the AI space, or will it encounter regulatory setbacks and competition from well-established tech behemoths?
One thing for sureβDeepSeek AI has already made its mark in the world of AI, and we will see what happens next closely.
Did you use DeepSeek AI? Tell us about it in the comments! Also, don’t forget to subscribe to our blog for more AI, business, and technology insights.
