Could China Win the AI Game?
.png)
As most media, businesses, law agencies, and consumers in the West focus on the AI rivalry between OpenAI and Google, a real competitor is making remarkable strides in the AI field in the East.
China has taken on the role of a formidable competitor, and AI has become the latest battleground. Once countries were battling to see who has more economic power, now we’re fighting to see who will render realistic AI fingers and develop more efficient hardware.
Regardless of the implications of this AI race on the already complex geopolitical landscape, China is achieving significant advancements. The scale and scope of these innovations prompt an important question:
Could China win the ongoing AI game?
Changing tides
While people in the West play with generative AI and carelessly laugh at Midjourney’s struggle with rendering fingers, China is increasing its AI capabilities in leaps and bounds.
Earlier this year, experts estimated that China was lagging behind the US by as much as three years in terms of AI advancement.
The big problem was that Chinese companies put out subpar AI models that were modified versions of existing Western models. As in the case of the company 01.AI which despite $1 billion in backing, developed AI solutions based on Meta’s LLaMA model.
Proud citizens of the Red Dragon freely admitted that their efforts in generative AI rely on underlying systems emerging in the States and acknowledged the issue.
What changed?
China started “conscripting”
As a communist country, China quickly recognized the power of such technology and poured massive amounts of money into the field. After all, the fruits of the labor belong to the people.
The country shifted gears and started working hard to improve its AI capability. Government support is accelerating the development of new technology and its integration with various industries domestically as well as abroad.
Reports show that China’s investment in the field is projected to reach $38 billion by 2027. Not surprisingly, over 60% of that investment is allocated to hardware to discontinue reliance on Taiwan and Nvidia.
Will the latest Chinese AI analog chips destroy Nvidia?
The influx of funds is already yielding results as China is developing a highly efficient analog chip specializing in vision tasks. The ACCEL chip utilizes analogy and photonic computing combined with a fine-tuned architecture that outperforms the Nvidia A100 over 3.7 times in image classification.
While Nvidia’s product is capable of reaching 1,248 TOPS (tera-operations per second) in vision tasks, the simulated version of ACCEL hits 4,600 TOPS.
It’s not just speed, though. In terms of accuracy, ACCEL reached up to 92,6% in time-lapse video recognition and ImageNet classification tasks.
Moreover, it’s also an optical system that uses non-electrical methods of transferring and encoding information, such as through laser pulses at a certain wavelength to extract visual data. This means it’s not as energy-hungry as digital GPU systems.
These advancements indicate that China’s ACCEL chip could potentially set new standards in AI vision tasks, challenging Nvidia's dominance and pushing the boundaries of what AI hardware can achieve.
Curious case of Kling
Hardware is not the only area in AI that China is now leading the game. Despite the fact you can’t access Kling outside of China, this AI video generation model is a true rival to OpenAI’s Sora.
The battle is on as Kling directly fired shots against OpenAI by copying the first third-party video generated with Sora.
The model was developed by Kuaishou Technology (TikTok’s main competitor in the East). At the moment, it can process text and turn it into 2-minute-long videos in 1080p resolution. Kling uses cutting-edge 3D variational autoencoders to reconstruct bodies and faces, allowing for a detailed expression and limb movement created from a single full-body image.
On top of that, the technology is enhanced with a 3D spatiotemporal joint attention mechanism that allows Kling to handle super complex scenes and movements. In other words, everything generated with this model looks like the real thing as videos accurately represent real-world physics laws.
Should you be afraid?
Should we even care about the fact China is starting to blow the West outside the water?
In a way, yes - because the AI war is not just about generating videos of cute pandas.
The United States is heavily invested in AI capabilities, with companies like OpenAI and Google leading the charge in advancements, particularly in natural language processing and machine learning. However, the focus tends to be on specific areas such as large language models (LLMs), raising the question of whether this constitutes a broader innovation landscape or a more narrow focus.
And what’s the situation in the EU, you might ask? Europe is currently stuck in the regulation game and is behind everyone else by focusing on regulating rather than innovating. This regulatory emphasis, while important, is slowing down Europe’s ability to keep pace with extremely rapid technological advancements seen in the US and China.
The situation is very complex with innovation in the US being led by corporates with their own set of private interests and Europe regulating instead of innovating. This leaves room for China’s latest investments in AI technology to potentially tip the odds in its favor. A monopoly in AI, especially by a country that has a different stance on personal freedoms, is a worrying prospect for Western citizens.
The “beast from the east,” as some might call it, could leverage artificial intelligence in ways that significantly impact global dynamics. For instance, the ability to dominate critical infrastructure and industries through advanced AI capabilities could lead to unprecedented control and influence. This potential for an AI monopoly surely raises concerns about privacy, security, and the balance of technological power, the way people in the West understand it.
Is China already winning?
The short answer is “who knows”. At the moment, the majority of the innovation in AI across both the West and the East is done behind closed doors. The tools released to the public are just the tip of the iceberg of immense research and data being poured into the development and only people behind closed doors know the real state of AI.
While in every industry competition is always welcome for it encourages progress, scale, and reduction in costs, the world has yet never seen a technology so powerful and so global growing at such an unprecedented rate. Pair it with the fact that AI technology can be easily misappropriated for nefarious uses, including military applications, cyber-attacks, scams, or mass disinformation, and we have a situation where everyone should really care who is leading and who is regulating the AI game.
So while a scenario where the US citizens fight the Chinese AI robots most likely won’t play out any time soon, these latest advancements do spell a sign of trouble in the overall AI landscape.
The question is, can the world keep up?