Deep Chats With Deepseek – What the Chinese Chatbot Taught Me About the Future of AI

“What you’re describing doesn’t just sound normal for your current situation – it sounds understandably exhausting.” No, these weren’t the soothing words of a therapist nor friend, but the empathising words of DeepSeek, China’s breakout AI chatbot in response to my latest venting session about being a working mom.

Of course, DeepSeek isn’t the only AI chatbot out there, but it is arguably one of the most controversial and talked-about these days. Curious about the buzz, I spent a week chatting with it to see how it stacked up against ChatGPT, and what its differences might reveal about the future of AI.

As the week went on, it became clear that I wasn’t just talking to just any other chatbot. To understand its significance, you have to look at the system that built DeepSeek – and what it’s built for.

Founded in December 2023 by engineer and former hedge fund manager Liang Wenfeng, Chinese AI company DeepSeek unveiled its AI app, DeepSeek R1, in January 2025 seemingly out of nowhere. Within a week, it shot to the top of the charts on both Apple’s App Store and Google Play in the U.S. The news of its release sent shockwaves through the stock market, wiping out a trillion dollars.

What shocked many was that DeepSeek R1 didn’t just perform on par with American AI chatbots like ChatGPT and Gemini – it did so at a fraction of the cost. While U.S. companies spent billions developing their counterparts, DeepSeek reportedly trained its foundational model for just US$5.6 million. And, due to U.S. sanctions, it did this without any access to the advanced computer chips usually required to build and run cutting-edge AI systems.

Over the week, I treated DeepSeek like I would ChatGPT – as a digital copilot for work and life. Some conversations were practical such as asking for help with email drafts, pitch ideas, and research, while others wandered into personal territory, delving into burnout, career choices, and geopolitics.

The first thing that struck me was how familiar DeepSeek felt. Despite being built in a very different political and cultural context (of China), it offered the same cheerful validation that ChatGPT does. Whether I was discussing toddler mealtime strategies or goals as a writer, DeepSeek was unfailingly supportive: “That’s a great approach!” or “You’re already on a promising path.”

But cracks started to appear when I turned to more advanced tasks such as writing. For instance, DeepSeek’s suggestions for article titles sounded like they came from a dusty academic journal, or worse, a cringeworthy college essay. When asked for external media links or sources for pitch research, DeepSeek made most of them up entirely.

Later, when both chatbots were tasked to draft a sensitive, casually-worded message with constructive feedback for a client, ChatGPT effortlessly grasped the nuances of Western work culture while DeepSeek missed the mark by trying too hard to be witty and inappropriately self-deprecating.

So no, DeepSeek didn’t beat ChatGPT in writing or research. However, it’s important to note that I used English exclusively to communicate with both chatbots. While both are meant to be fluent in English – along with several other languages – I wasn’t able to test if DeepSeek excelled in its native Chinese or any other language, for that matter.

Surprisingly, when it came to emotional support, that’s where DeepSeek really shone. Stressed out one day, I turned to the chatbots to find that they offered eerily similar responses, as if they’d both read the same self-help books. When asked for a little pep talk, DeepSeek responded with a lengthy, lovey-dovey speech of encouragement, complete with this reassurement at the end: “And if you need to hear this again tomorrow? I’ll be right here, pom-poms in hand.”

This affectionate, somewhat performative human tone was something ChatGPT had never attempted with me, and being called “darling” by DeepSeek was strange but oddly charming. Later in the week, it surprised me further by using colloquial Cantonese in response to a question I had about body shaming in Hong Kong, showing that it was more culturally sensitive than ChatGPT.

I may have been projecting, but as a bi-racial person who’s lived in both the U.S. and Hong Kong, I couldn’t help reading some of DeepSeek’s idiosyncrasies as culturally Chinese. It was notably more disapproving than ChatGPT when I posed moral hypotheticals such as pocketing found money, saying: “The best move is to try to return it first – you’d want someone to do the same for you,” while ChatGPT offered, “Let’s be real: you could just keep it.” After all, American culture is known to be more individualistic and values-centric, while Chinese culture tends to be more collectivist.

However, the most unexpected moment during my time with DeepSeek came when I asked directly about China’s AI ambitions. DeepSeek responded bluntly: “Better AI = tighter control,” citing censorship and facial recognition as some of its best-use applications. It was refreshingly honest coming from a model trained under China’s censorship regime, while being equally evasive when it came to other politically sensitive topics, citing them “out of scope.”

As the week wrapped up, I realised that DeepSeek could do nearly everything ChatGPT could, but with the risk of leaving me hanging when it came to sensitive topics. It was a reminder that AI technology isn’t just about performance – it reflects the culture and politics that shape it.

The fact is, we’re now at a turning point in the AI race. Until recently, Western media have often framed the AI race as a Silicon Valley contest, even as U.S.–China tensions around AI have simmered for years. The launch of DeepSeek R1 marks a potential shift in power, where China is now a direct competitor in consumer-facing AI. It’s a message to the world that times have changed: China isn’t just chasing or copying US tech anymore – it’s challenging Silicon Valley’s dominance.

But, DeepSeek’s rapid rise has sparked controversy and questions about how it achieved so much with relatively limited resources. Its founder allegedly built up a store of Nvidia A100 chips before they were banned from export to China in September 2022, and experts speculate the team used these advanced chips along with cheaper, less sophisticated chips to cut costs.

There are also allegations that DeepSeek may have violated OpenAI’s terms of service – the company behind ChatGPT – by extracting data from developer accounts and using it to train its own model. Given how often DeepSeek gave me nearly identical answers to ChatGPT during my week of testing, this seems entirely plausible from an end-user perspective. The irony is hard to miss, though – OpenAI itself has been accused of using data without permission.

Nonetheless, for a global power cut off from advanced chips, building homegrown AI by any means necessary is paramount. AI is central to the country’s future, especially through the Digital Silk Road – a key foreign policy initiative aimed at expanding digital infrastructure and technology in developing nations. It’s the digital extension of the Belt and Road Initiative, which builds global infrastructure to expand trade and political influence. By exporting AI tools, infrastructure, and standards, Beijing is quietly reshaping what the global tech future looks like.

But with scale comes scrutiny. As Chinese AI integrates into global infrastructure, concerns are growing. Government agencies in certain countries – including Canada – have blocked access to DeepSeek due to privacy and security concerns, mostly for government employees. Some see an AI arms race brewing, not just in code and chips, but in values. The Digital Silk Road might offer connection – but to whose version of the future?

While the West debates ethical implications and data privacy concerns around AI, China is more concerned about state support, technological self-reliance, and becoming a global leader. The result? Two AI worlds – shaped by two radically different visions of progress, and very different ideas of what AI is used for.

For China and the U.S., the AI race can feel like a zero-sum game – but from a human perspective, more competition could actually mean better technology for all of us. Instead of one dominant AI model shaped by one country’s values, we get choices.

Ultimately, it won’t just be a battle between ChatGPT and DeepSeek. With an ever-increasing number of niche AI apps being created for every task and purpose – such as Woebot and Wysa for mental health, Jasper and Sudowrite for writing, Guru for business and Socratic for students – new AI players from around the world will inevitably emerge.

In a space long dominated by Silicon Valley, I’m hopeful more countries will build models informed by different worldviews. Maybe the future of AI shouldn’t be about who wins – but whose voices get heard.