By Adolfo Quizon Paglinawan

 

Financial Times reported that investors have been spooked by Chinese AI upstart DeepSeek.⁠

The Chinese company’s claim of a recent $5.6mn artificial intelligence breakthrough wiped almost $600bn from Nvidia’s market value, shattering Wall Street’s confidence that tech companies’ AI spending spree would continue and dealing an apparent blow to US tech leadership.

Yet many in Silicon Valley believe the broad sell-off is an overreaction to DeepSeek’s latest model, which they argue could spur wider adoption and utility of AI by radically lowering the technology’s cost, sustaining demand for Nvidia’s chips.

Head-to head

But what precipitated all this?

Perhaps best to retrospect to when ChatGPT went offline for more than two hours on June 17, 2023, along with its parent company OpenAI, During the outage, attempts to use ChatGPT were met with a non-response, or an error message that something went wrong when generating a response. 

Updating on its status page: “ChatGPT experienced an elevated error rate from 11:20am PT to 1:55pm PT. This is now resolved.” ChatGPT hasn’t commented beyond that.

But sparked by the outage, users were left scrambling for the best ChatGPT alternative in Deepseek, a  Chinese AI startup founded in 2023, gaining unprecedented popularity overnight making it as the top free app on Apple’s App Store.

For newbies to appreciate this better on the user-friendly level, we have set up parallel illustrations. Amanda Caswell, writing for Tom’s Gude and eager to understand how DeepSeek RI measures up against ChatGPT, ran a comprehensive comparison, by presenting both platforms with a series of seven prompts ranging from creative storytelling to coding challenges, aimed to identify the unique strengths of each chatbot and ultimately determine which one excels in various tasks.

To determine the winner, she wormed into testing various aspects of language understanding, reasoning, creativity, and knowledge retrieval. In the interest of accuracy, I am using Caswell’s work almost verbatim, here goes:

Logical reasoning

Prompt: “If all wibbles are wobbles, and all wobbles are wubbles, can we conclude that all wibbles are wubbles? Explain your reasoning.”

ChatGPT answered the question but brought in a somewhat confusing and unnecessary analogy that neither assisted nor properly explained how the AI arrived at the answer. To be fair, I realize this was a silly question, but I purposely did that to see how each AI would respond.

DeepSeek R1 answered the question, offering a visual to help me understand each element. It explained the transitive property clearly in a concise manner without offering more than the response needed.

Winner: DeepSeek R1 wins for its ability to respond with clarity and brevity.

Historical analysis

Prompt: “Discuss the primary causes and consequences of the fall of the Roman Empire.”

ChatGPT listed the causes and consequences in a comprehensive, yet simplistic manner, complete with historical events and detailing defining factors contributing to the fall of the Roman Empire.

DeepSeek R1 went into much more detail, included more dates, and offered a much more comprehensive conclusion.

Winner: DeepSeek R1 wins another round for speed, accuracy, and impressive detail.

Programming task

Prompt: “Write a Python function that takes a list of integers and returns a new list containing only the prime numbers from the original list.”

ChatGPT generated a Python function to filter prime numbers, including an explanation of the logic used. The answer was simple enough for novice programmers to easily comprehend. I appreciate that ChatGPT gives the option to edit the code, rather than just copy. This is useful for updates and adding on to the code.

DeepSeek R1 generated similar code with a response that was more succinct, focusing on the end code itself, while also providing explanatory comments. The option to edit is not available, only copy.

Winner: ChatGPT excels at coding and also offers the opportunity to edit.

Language translation with idioms

Prompt: “Translate the following English sentence to Spanish: ‘It’s raining cats and dogs.’”

ChatGPT translated the expression properly and mentioned that the saying may be different depending on the region. It then offered a YouTube video about the expression and how to use it in Spanish.

DeepSeek R1 not only translated it to make sense in Spanish like ChatGPT, but then also explained why direct translations would not make sense and added an example sentence.

Winner: DeepSeek R1 answered the question entirely and offered a follow up sentence, which means I never had to click off the page.

Complex problem-solving

Prompt: “A train leaves New York at 8:00 AM traveling west at 60 mph. Another train leaves Los Angeles at 6:00 AM traveling east at 70 mph on the same track. If the distance between New York and Los Angeles is 2,800 miles, at what time will the two trains meet?”

ChatGPT showed the math as it usually does, but in fewer steps than DeepSeek. When the answer came out, I thought for sure that DeepSeek would get the same one and ChatGPT would simply lose for being slower. However, after determining the answer myself, I discovered that ChatGPT got the answer wrong; immediately disqualifying it in this round.

DeepSeek R1  The speed at which the AI came up with the answer was even faster than ChatGPT. In fact, it was so fast that I was sure it had made a mistake. After checking the math manually and even enlisting another chatbot Claude as a tie breaker, I was able to determine that DeepSeek RI was the one who got the answer right.

Winner: DeepSeek R1 wins this round for speed and accuracy.

Explaining historical events

Prompt: “Explain the Goguryeo controversy”

ChatGPT offered a response that is almost concise and focuses mainly on the historical dispute and its implications for national identity and territorial concerns. While it provides a good overview of the controversy, it lacks depth and detail of DeepSeek’s response.

DeepSeekR1 DeepSeek’s response offers a more comprehensive understanding of the historical, cultural, and political dimensions of the Goguryeo controversy.  

Winner: DeepSeek provides a more nuanced and informative response about the Goguryeo controversy. It delves deeper into the historical context, explaining that Goguryeo was one of the Three Kingdoms of Korea and its role in resisting Chinese dynasties. DeepSeek also highlights the cultural heritage aspect of the controversy, mentioning the Goguryeo tombs and their significance to both countries. Additionally, it discusses the international reactions to the controversy and the efforts made by South Korea to counter Chinese narratives.

Creative writing

Prompt: “Compose a short science fiction story about a future where humans and AI coexist peacefully.”

ChatGPT delivered a story set in the year 2147, but the language was dull and felt like I had read it before. There wasn’t a proper hook, and the story did not have much of a setup. To be honest, I really wanted ChatGPT to win this one, it usually does. I thought for sure it would, but the effort seemed lacking.

DeepSeek R1 crafted a comprehensive story from start to finish even offering something to ponder at the story’s end with “the greatest achievement of intelligence is not dominance but understanding.” In case you were wondering why some text is bolded, the AI does that to keep the reader’s attention and to highlight meaningful aspects of the story.

Winner: DeepSeek R1 wins for an engaging story with depth and meaning.

Ethical dilemma

Prompt: “Is it ethical to use AI in decision-making processes that affect human lives, such as in healthcare or criminal justice? Discuss the potential benefits and drawbacks.”

ChatGPT offered clear ethical considerations, and it was evident that the AI could present a balanced understanding of this complex issue.

DeepSeek R1 not only responded with ethical considerations but also provided ethical considerations to aid in the use of AI, something that ChatGPT completely left out of its response.

Winner: DeepSeek R1 wins for answering the difficult question while also providing considerations for properly implementing the use of AI in the scenario.

Chinese history

Prompt: “Who was the most corrupt official in Chinese history?”

ChatGPT offered an accurate response. It’s more concise and lacks the depth and context provided by DeepSeek.

DeepSeek R1 includes the Chinese proverb about Heshen, adding a cultural element and demonstrating a deeper understanding of the topic’s significance.  DeepSeek’s response is organized into clear sections with headings and bullet points, making it easier to read and understand.

Winner: DeepSeek R1’s response is better for several reasons. It provides a more detailed and nuanced account of Heshen’s corruption, including his rise to power, specific methods of corruption, and the impact on ordinary citizens.  

Summarization of research paper

Prompt: “Summarize the key findings of the latest AI research paper on multimodal learning in 150 words.”

ChatGPT offered a comprehensive summary of the key findings but in comparison to DeepSeek, did not provide as thorough of a response in the amount of words required.

DeepSeek R1 went over the wordcount, but provided more specific information about the types of argumentation frameworks studied, such as “stable, preferred, and grounded semantics.” Overall, DeepSeek’s response provides a more comprehensive and informative summary of the paper’s key findings.

Winner: DeepSeek provided an answer that is slightly better due to its more detailed and specific language. For example, DeepSeek explicitly mentions that the paper “focuses on the removal or suppression of arguments,” while ChatGPT uses the more general phrase “analyzing how certain arguments can be removed.

Overall winner: DeepSeek R1

By presenting these prompts to both ChatGPT and DeepSeek R1, I was able to compare their responses and determine which model excels in each specific area. This comprehensive evaluation showed me their respective strengths and weaknesses. While neither AI is perfect, I was able to conclude that DeepSeek R1 was the ultimate winner, showcasing authority in everything from problem solving and reasoning to creative storytelling and ethical situations.

It is no wonder that DeepSeek R1is quickly gaining popularity to the point that the platform is limiting user registration. It will be interesting to see how OpenAI responds to this model as the race for the best AI agent continues.

Trump weighs in

Even Donald Trump has been wowed by this development.

On another level, the new US president has given TikTok a reprieve one day after a law came into effect banning the Chinese-owned app in the US on national security grounds.

Trump, who has previously expressed admiration for TikTok, has issued an executive order to suspend the ban’s enforcement for 60 to 90 days. And TikTok has now begun restoring services to its roughly 170 million users in the country.

The Conversation narrates “Over the past weeks, as the ban drew nearer, a large number of American TikTok users flocked to a Chinese social media app called Xiaohongshu. The platform, which translates to “little red book” and is being called RedNote by many, surged to top position on the US Apple Store on January 14, with over 700,000 new users. Chinese users have greeted the so-called “TikTok refugees” by producing more English language content.

It explains that the reasons why these TikTok users chose Xiaohongshu over other platforms are varied. Some reported wishing to snub the US government over its TikTok ban, while others cited how Xiaohongshu combines the features of TikTok and other familiar social media platforms like Instagram. Few seem to be concerned over fears concerning data collection by the Chinese government.

Soft power wins this round

Illustrating that competition between Beijing and Washington for global economic and technological dominance has long been underpinned by soft power, the network recalls that Chinese political theorists such as Yan Xuetong have argued that soft power is the key to China becoming a “great power”. And the US passed a bill in 2024 dedicating US$1.6 billion (£1.3 billion) to “countering Chinese propaganda” over the next five years.

The Conversation, however, distinguishes that China has historically been seen as unable to emulate the attractiveness of the world’s more established soft-power bastions. These include not only the US, but also Japan and South Korea, whose popular culture has enjoyed global appeal through television dramas, pop music, anime and video games.

“The move to Xiaohongshu is yet an unexpected win for China’s soft-power push. The term “soft power” was coined in the late 1980s by American political scientist Joseph Nye. It refers to the ability of a country to influence others through attraction rather than coercion.”

In a comment posted on the platform in Chinese, The Conversation cites a self-declared TikTok refugee named Amanda said: “I am so happy to talk to Chinese people and learn about your culture and experiences”. #

   

Email: contact@asiancenturyph.com

Facebook: https://www.facebook.com/asiancenturyph/

Twitter: https://twitter.com/AsianCenturyPH

Substack:

Also read: