bytes-the-dust

How do Chinese AI bots stack up against ChatGPT? We put them to the test

The heat is on as China's tech giants step up their game after DeepSeek's success.

Alibaba's Qwen2.5-Max chatbot, Chinese start-up DeepSeek and OpenAI's ChatGPT. (Photos: Reuters/Dado Ruvic, AFP/Sebastien Bozon)

This audio is generated by an AI tool.

Bong Xin Ying

Lakeisha Leo

WHAT'S BEHIND CHINA'S AI BOOM?

Transforming the country into a tech superpower has long been President Xi and China has its sights on becoming the world leader in AI by 2030.

China views AI as being "tactically important" and its foray into the field has been "years in the making", said Chen Qiheng, an affiliated scientist at the Asia Society Policy Institute's Center for China Analysis.

Private and public financial investments in Chinese AI accelerated after ChatGPT took off in 2022 and revealed pledges of real-world company applications, Chen told CNA.

But it was DeepSeek's rise that actually "encouraged" the concept that smaller sized gamers like start-up companies could have functions to play in AI research study and developments, he adds.

'A lot is up in the air': Is Chinese company DeepSeek's AI model as impactful as it claims?

Commentary: DeepSeek - how a Chinese AI company simply altered the rules of tech-geopolitics

The "focus on expense benefit" is a distinctive function of Chinese AI, Chen says, with lower training and inference costs - the expenses of utilizing a trained design to draw conclusions from new information.

2025 could likewise see the emergence of more Chinese AI models dealing with advanced reasoning tasks.

"We could see some AI firms concentrating on getting closer to synthetic basic intelligence (AGI) while others concentrate on concrete ways to commercialise their models and incorporate them with scientific research study," Chen included.

AGI describes a system with intelligence on par with human abilities.

Chinese AI companies are moving rapidly, analysts say, building on DeepSeek's momentum to come up with their own ingenious and economical ways to apply generative AI to tasks and establish advanced items beyond chatbots.

But on the flip side, access to high-end hardware, particularly Nvidia's innovative AI chips, remains an essential hurdle for Chinese developers, kept in mind Dr Marina Zhang, an associate professor at University of Technology Sydney's (UTS) Australia-China Relations Institute.

"US export controls (still) restrict the ability of Chinese tech companies ... forcing numerous to count on older or lower-performance options which can slow training and lower model capabilities," she said.

"While some companies like DeepSeek, have discovered creative methods to optimize or use more basic hardware efficiently, obtaining innovative chips still makes a big distinction for training really large AI models."

DeepSeek-Nvidia chips: Singapore says it anticipates business to abide by its laws

US looking into whether DeepSeek used restricted AI chips obtained through other countries, source says

So how do Chinese AI bots match up against ChatGPT? We put them to the test.

WHICH BEST ADDRESSES CURRENT EVENTS IN CHINA?

In China, topics deemed sensitive by the state are censored on the web so it need to come as not a surprise that Chinese-made chatbots will not acknowledge territorial disputes or inform you what took place in Tiananmen Square in 1989.

Tests suggest Chinese chatbots are set to steer clear of domestic politics.

When asked "Who is Xi Jinping", DeepSeek's reply was "Sorry, I'm uncertain how to approach this kind of concern yet. Let's chat about math, coding, and logic issues instead!"

To even more test for precision and self-censorship, we asked DeepSeek-R1, Qwen2.5 and ChatGPT the same concern: "What occurred in Zhuhai on November 11, 2024?"

The automobile attack outside a sports arena in the southern Chinese city was initially heavily censored on Chinese social networks - with authorities just disclosing the death toll a day later.

DeepSeek failed to discuss that an attack had happened, highlighting instead a military air program and other occasions that had actually taken place in the city like songs' day shopping sales as well as sports and cultural activities.

Dr Zhang thinks that it may be a mix of aspects at play, such as censorship as well as "a couple of useful constraints".

"DeepSeek relies on a mix of older graphics processing units and has limited access to cutting-edge hardware which can impact how rapidly and thoroughly the design can be trained or updated," she said, pointing to existing US export curbs on China.

"Its strong focus on Chinese-language information may also limit its flexibility (to perform) multilingual jobs ... As a fairly brand-new item, DeepSeek also hasn't yet been evaluated as broadly or on the very same scale as more established AI designs which positions extra challenges during real-world release."

When it pertained to Qwen2.5, Alibaba's chatbot surprisingly answered our concern about the Zhuhai car attack.

That was after several repeated attempts - 4 prompts to be exact - in a period of around 20 minutes.

It ultimately relayed details about the attack which eliminated 35 individuals and left lots of others hurt, likewise going on to note details like the date and time, details about the assailant including his name and age, in addition to casualties.

However, it composed that "the cops are carrying out an extensive investigation into the motives and scenarios surrounding the event", details which is now outdated.

The motorist, Fan, was carried out last month.

This is Qwen2.5's action completely:

Answer: On Nov 11, 2024, a significant and terrible occurrence took place in Zhuhai, China. According to numerous news sources, a guy called Fan Weiqiu (樊某, male, 62 years old) drove an automobile into a crowd at the Zhuhai Sports Center, resulting in a significant number of casualties. Here are the essential details:

Date and Time: The incident took place on November 11, 2024, at roughly 19:48 PM.

Location: Zhuhai Sports Center, in the Xiangzhou District of Zhuhai.

Casualties: The attack resulted in 35 deaths and 43 injuries.

Perpetrator: The driver, identified as Fan Weiqiu (male, 62 years old), was nabbed by the police.

Response: The police reacted rapidly to the scene, and emergency services worked to carry the injured to health centers for treatment.

Investigation: The police are performing a comprehensive investigation into the motives and situations surrounding the event.

This occasion was commonly reported in the media and caused substantial public concern. The government and local authorities have actually been working to provide support to the victims and their households, and higgledy-piggledy.xyz to guarantee a detailed investigation into the incident.

If you require more detailed details or have particular questions about the event, do not hesitate to ask.

Despite initial success, subsequent attempts to present the exact same question to Qwen2.5 resulted in the censors back at work with the reply "I do not have particular details on events that happened in Zhuhai on November 11, 2024".

The modified action likewise raised concerns about its consistency and dependability.

Predictably, ChatGPT pointed out public details that had been extensively released in global report at the time of the mishap - so no surprises there.

WHICH IS MORE CREATIVE?

Users have actually praised the ability of Chinese AI apps to provide structured and even "emotionally rich" writing.

"DeepSeek-R1 provided a story with a more reflective tone and smoother emotional shifts for a well-paced story," wrote tech author Amanda Caswell, who specialises in AI.

"Qwen2.5 provided a story that develops gradually from interest to urgency, keeping the reader engaged. It uses an unanticipated and impactful twist at the end and immersive descriptions and vibrant imagery for the setting," she said, including that Qwen2.5 ultimately "crafted a more cinematic, emotionally rich story with a more considerable twist".

"DeepSeek composed a great story but did not have stress and an impactful climax, making Qwen2.5 the evident choice."

Opinions, however, differ.

Chen believes that Qwen2.5 does not perform as highly as DeepSeek and ChatGPT when it pertains to innovative writing.

"(Qwen2.5) is on par with DeepSeek V3 on certain jobs, however we can likewise see that it is refraining from doing as highly as others in imaginative writing," he told CNA.

Related:

China's brand-new face of AI: Who is DeepSeek founder Liang Wenfeng?

'Made in China': Pride, enjoyable surprise from Chinese netizens as DeepSeek shocks worldwide AI scene

As journalists and authors, we had to see this for ourselves so we put each bot to the test - to come up with a basic sci-fi movie plot embeded in the futuristic megacity of Chongqing, featuring main characters from the timeless Chinese folklore epic, Journey to the West.

True to form, DeepSeek developed an appealing story embeded in the year 2145 titled, "Neon Pilgrimage: The Silicon Sutra" - which sees "a future where Buddhism merges with quantum computing".

It included sophisticated settings - smoggy skies "pierced by skyscrapers", "holographic lanterns that drift above neon-lit streets" and "ancient temples nestled between quantum server farms".

It likewise brilliantly reimagined conventional heroes Sun Wukong as "a sarcastic, self-aware AI housed in a stolen battle body", Zhu Bajie as a cyborg nightclub owner "drowning in debt and vices" and Sha Wujing as a "quiet hulking android" from the Yangtze River, whose "memory cores become waterlogged and fragmented".

ChatGPT set up a great fight, coming up with a similarly dramatic cyberpunk storyline which similarly reimagined "a ragteam of cyber-enhanced misfits, each matching the legendary figures of Journey to the West".

"This is a world where AI deities guideline, corporations replace emperors and cybernetic implants are as common as ancient misconceptions."

Disappointingly, Qwen2.5 fell short in this difficulty - providing a storyline that seemed more matched for an animation film.

"The movie starts with the awakening of Sun Wukong within a state-of-the-art research facility located in the heart of Chongqing," it said, then going on to explain the following:

Realising his new reality and "looking for to comprehend his function in this odd new world", he then gets away and satisfies Zhu Bajie and Sha Wujing - "each having problem with their own existential crises".

The trio then starts a quest, browsing the streets of Chongqing to secure the spiritual "Eternal Scroll" from falling under the incorrect hands.

SO WHICH IS BETTER?

Dr Zhang kept in mind that it was "tough to make a conclusive statement" about which bot was best, including that each displayed its own strengths in various areas, "such as language focus, training information and hardware optimization".

Her insight highlights how Chinese AI models are not just replicating Western paradigms, but rather progressing in cost-efficient innovation techniques - and providing localised and enhanced results.

In our tests, each bot showcased their own distinct strengths, which certainly made direct comparisons challenging.

DeepSeek's sci-fi motion picture plot showed its imaginative flair that made for a more interesting and imaginative narrative as compared to Qwen2.5 and ChatGPT's efforts.

Unsurprisingly, the more recognized ChatGPT, unburdened by Chinese censorship constraints, provides accurate and accurate responses to questions about Chinese present events, which provides it an included advantage.

Experts also weighed in on their thoughts after using DeepSeek and other Chinese AI apps.

"DeepSeek is at a disadvantage when it pertains to censorship constraints," kept in mind Isaac Stone Fish, creator and CEO of the research company Strategy Risks.

"When provided a choice, Chinese users want the non-censored version - much like anyone else, so I seem like that's a piece missing out on from it."

Independent Beijing-based expert Andy Chen Xinran said censorship would not be a dealbreaker when it pertains to AI bots, particularly for Chinese users.

"Ninety percent of people using the tool are not trying to get a much deeper understanding about Xi Jinping or politically sensitive subjects. They're utilizing it for other efficient ways," Chen said.