ChatGPT’s Golden Age Is Over – Is China ahead in the technology revolution?

Image generated by OpenAI’s ChatGPT

The public’s first exposure to AI systems, for the average person, is learning about it either on The New York Times while drinking their morning Americano, or watching their X (Twitter) feed being flooded with #SamAltman. It amazes me that AI can do so much, as well as how much it’s developed over the course of the year or two that it has been in the market. For the past few months that GPT-4 has been prominent in public discussion, ChatGPT has been the monopoly, where the publicity of Google’s Gemini AI was turned away after it had generated images of Asian popes and Indigenous (culturally stereotyped) George Washingtons in a disastrous DEI failure, and other AIs like Claude had not been fully developed. But ChatGPT’s spotlight on the runway is about to be stolen, because DeepSeek had just entered with a red-and-yellow bamboo hat that the audience observed to be much more intricate than America’s simple red-white-and-blue baseball cap, on which some say they could see a faint MAGA stitching. The real question is: will ChatGPT allow DeepSeek to steal its spotlight (not to mention the stability of the entire US stock market)?
DeepSeek stands by its name. Despite having little marketing exposure in North America, it quickly rose to popularity in Asia, in which the echo chambers of social media quickly bumped it to the US. This is mainly because DeepSeek was better than ChatGPT in almost all the features that Altman prided his platform on, diving “deeper” in almost every response. Since then, ChatGPT has been racing to innovate, but has been extremely slow in doing so.

The Chinese platform showed the world that it could provide equal quality of service, but at a much faster rate and vastly lower cost. The goal of DeepSeek is to create artificial general intelligence, and the company’s improvements in AI reasoning skills mark a major step forward in technological development. Unlike ChatGPT, which was trained on large amounts of data, DeepSeek was trained on logical reasoning. It built itself on the only partially patented GPT system, where whatever data needed was provided. However, the major difference is that ChatGPT needs to scour its colossal amount of training data for one statistic, whereas DeepSeek separates its data into categories and looks through the data set of a specific category based on the user’s prompt. 

This feature is not only more efficient, but it also reduces the environmental impact of AI. GPT requires large amounts of water to cool its servers and systems – approximately an average of a standard 500mL bottle of water per user prompt. It affects the developing countries that are already in need of clean water, where these servers are based. Many environmentalists and humanitarians alike criticize ChatGPT for its intensive water use and have started to boycott the company. 

Yet, the most important aspect of these AI technologies is not their training or water usage – but the chips that they (specifically ChatGPT and DeepSeek) use. DeepSeek proved that it could use a significantly lower number of semiconductor chips to achieve higher-quality responses. From when ChatGPT was first introduced to when DeepSeek started to rise on the global stage, Nvidia stocks were booming. The most widely accepted explanation for Nvidia’s stock price increase is that the company is the driving force behind the AI revolution. Millions of GPUs are necessary for the AI revolution, and Nvidia sells a large number of them, earning hundreds of billions of dollars. Investors believed that the AI industry would heavily rely on a large number of chips, but this notion has now been disproved by DeepSeek. Therefore, many deemed it unprofitable to invest in a stock that will not, in fact, reshape the tech world. So, Nvidia wiped around a trillion dollars off the market capitalization of America’s listed tech firms, sending a blood pressure spike to the entirety of Wall Street.

Nvidia Corp share price over the past month, dropping to an all-time low.

For the average consumer, the real question is whether DeepSeek will take over ChatGPT. China’s AI follows the country’s stereotypical business model: cheap, fast, and effective. Upon experimenting with both, and even subscribing to ChatGPT Plus for a short period of time (and, of course, canceling it after discovering DeepSeek), I found the new AI’s DeepThink feature to be the most helpful. The DeepThink feature walks users through the entire thought process of the AI, where one can see it eliminating possible confusing answers or irrelevant information, fine-tuning the logic of its response masterfully. I’ve observed that it’s generally good at math (aka, that DeepSeek can solve the grade 12 math contest problems that ChatGPT cannot), because it was designed based on logical reasoning. 

This article may seem like a glorification of AI in general. But I must acknowledge the limitations of AI. A common experiment I found somewhere in the New York Times last year was asking AI how many letters are in a word – or, simply, asking how many of one letter are in a word. So, I asked ChatGPT “How many Rs are in the word strawberry”? It responded with an innocent “There are two Rs in the word strawberry. Let me know if you need help with anything else!”. 

There are, in fact, three Rs in the word strawberry. 

DeepSeek, however, was much more logical and clear in its approach. Instead of the one-liner response delivered to me by ChatGPT, this new AI sent me a seven-step methodical approach for finding the number of Rs in the word strawberry, including listing out all the letters, analyzing each one and even double-checking. The difference between the processes of producing these responses is that ChatGPT is a numerical-based model, translating all words inputted by the user into numbers that correspond to each word, and next analyzing its specific meaning. It then ties the numbers together into a response, and finally translates its numerical response into words. So, ChatGPT does not truly know words, nor can it decipher how many letters there are in one. DeepSeek, however, is entirely vocabulary-based. But who will take over the world? The answer is neither. I’m sure that there are already at least two movies being filmed about an AI popping out of a computer screen and taking over the Earth. But we need to break free of these sci-fi myths, because it will be a long time before we can truly build a dystopian AI world that will lead to the plot of The Terminator.