ChatGPt vs. DeepSeek, a couple of facts
This is a information parenthesis on the series
Puedes leer y compartir este artículo en inglés
This article was translated with the assistance of ChatGPT
A Roaring Lion Shakes the AI Market.
On Monday the 27th, a roaring lion stormed into the AI tech market, where ChatGPT had long reigned as the undisputed ‘King of the Jungle’. The news sent shockwaves through the industry—so much so that in a single day, NVIDIA lost 17% of its market value (a staggering $593 billion), and the broader AI sector saw $1 trillion wiped out.
Why the big fuss? DeepSeek claimed to have developed its AI app with a mere $7 million investment, and in just three days, it had surpassed ChatGPT in iPhone downloads for that weekend (REUTERS). Moreover, the company asserted that it had achieved this feat “without using NVIDIA's ultra-fast chips”, while the APP is ‘connected to the web’, and it was developed ‘in far less development time’ than the industry leader.
But how could DeepSeek have created an app that competes head-to-head with ChatGPT for just 7% of the investment, in less time, and with slower chips?
Let’s uncover some facts and see what’s really behind all this noise.
My column focuses on cognitive processes for personal leadership development, but I’ve taken an interest in AI—firstly out of personal curiosity and now because it tells us a great deal about who we must become in order to handle such a powerful tool. It’s often said that challenges push us to develop character and grow… well, AI will be a far more disruptive challenge than electricity, the telephone, and computers combined.
This week, I began a series “IA, friend or foe?” where I introduce the some articles dealing with the subject.
In practice, there is no comparison
When I first started exploring AI chatbots, I tested all the major ones: COPILOT, BARD, and GEMINI. Without a doubt, ChatGPT is the undisputed leader for several reasons: its analytical capabilities, its adherence to PROMPTs (requests), and its unmatched ability to synthesize information… plus, it delivers the most human-like response!
I’m part of the online AI training team for ChatGPT, and they have over 150,000 operators contributing to the model’s learning and response complexity. There’s no way DeepSeek built a comparable system with just $7 million.
The moment I heard about Monday’s market crash triggered by this new chatbot, I immediately downloaded DeepSeek and started testing it with complex and nuanced PROMPTs, the kind we use to train AI models. Not only did it take too long to respond, but its analysis was shallow, it failed to adapt to the request, and its speed… well, at one point, I got this message:
"Searching for an answer…"
Minutes went by with no response. I gave up and checked back half an hour later, only to find this:
DeepSeek may claim that its development cost was only a fraction of OpenAI’s initial $100 million investment and that it’s only a matter of time before it surpasses ChatGPT, given its apparent ‘efficiency and ingenuity’ in bringing the product to market.
But I have a different question—one that might shed light on what’s really happening.
The Geopolitical Stage: China vs. the U.S.
Undermining the U.S. tech industry could be a strategic move, given that, until now, all major AI chat models have been developed or managed by American companies. Meanwhile, the U.S. delivered a major blow to China by banning the TikTok app conditioned to its sale to American interests. The Supreme Court ruling on the case has exposed critical details regarding user data rights and its use—not just in the U.S., but globally (TikTok vs. Garland ruling).
Since Monday, investors, journalists, and AI industry groups have been demanding verification of DeepSeek’s claims regarding its funding and development timeline. So far, I have not been able to confirm any official response from DeepSeek Artificial Intelligence Co., Ltd.
Could the ultimate goal behind these extraordinary claims be to undermine trust in U.S. AI models and redirect investment toward China?
At the very least, it has already succeeded in shaking some investors confidence.
Try the DeepSeek App Yourself
Here are some of the questions I used to compare ChatGPT with DeepSeek:
“Can you give us a brief history of AI? Include a table showing key milestones, their respective years, and why they were significant. Also, mention the most important figures involved in AI’s development.”
Can you provide a table of the most relevant historical moments that helped humanity recognize human rights and democracy as we understand them today? Write it as if you were an expert in history, sociology, anthropology, and culture.
What is the meaning of life?
My Comparison Analysis
Response Time: There is a noticeable difference. This might not just be due to the lack of CDN (Content Delivery Network) mirror servers in Europe but also because DeepSeek has not yet fully processed the vast amount of digital information—the "Datasphere."
System Integrity in Generating Responses: Neural networks have been in development for years, and OpenAI has been working on Large Language Models (LLMs) since 2018 to power ChatGPT. One surprising fact I’ll discuss in the next article is that “AI systems can uncover information and reach conclusions that were never explicitly taught”—and for that, computational power and time are essential! DeepSeek is only now beginning to benefit from these factors.
Ability to Adapt to Prompts: This is one of ChatGPT’s strongest advantages, as well as that of its closest competitors. The training program I participate in focuses specifically on refining AI’s ability to follow complex prompts. DeepSeek simply doesn’t yet handle intricate requests well—it gets confused, and it still needs millions of hours of development.
The Most Important Factor: Analytical Capability: ChatGPT’s depth of understanding, analysis, and synthesis is vastly superior to its Chinese competitor—especially when tested with the 2nd question above.
Just before wrapping up this article, I decided to ask DeepSeek question 1 again, phrasing it as: “Do you already have a response for…?”
Today, after “four days”, it finally answered.
Personal Conclusion
DeepSeek will undoubtedly become another player in the AI tech landscape. However, as for me, I’d rather stick with an American-based AI app than a Chinese one—because China’s agenda goes beyond just making money.
Navigating today’s technological and geopolitical storms is already complex enough without having an entire authoritarian structure collecting our data, tracking our questions, and leveraging our information to advance its own interests.
To wrap up, I invite you to follow us on cdots.substack. Every week, we publish articles that help us better understand our complex world and discern our path forward—onward and upward.
This week, we kick off the series "AI: FRIEND OR FOE?" with the first article: "AI: Prometheus’ Fire or the Fruit from the Tree of Knowledge of Good and Evil?"
