The year 2024 was a remarkable one for China’s large model industry, marked by significant events that reshaped the landscape of artificial intelligence. Here’s a recap of the top ten events that left a lasting impact.
01. Artificial Intelligence+ Integrated into the Government Work Report
In March, the “Artificial Intelligence+” initiative was incorporated into the Government Work Report for the first time, elevating AI to a strategic level. Predictions suggest that by 2035, generative AI could contribute nearly 90 trillion yuan to the global economy, with China accounting for over 30 trillion yuan.
This move reflects a profound insight into the development of technology and a forward-looking layout, providing policy support and directional guidance for the advancement of large models.
02. Kuaishou’s KeLing AI Ignites the Visual Generation Track
On June 6, Kuaishou entered the fray with its self-developed video generation model, KeLing AI. Since its launch, KeLing AI has undergone nearly 20 product functionality upgrades. By December 10, 2024, it had amassed over 6 million users.
The emergence of KeLing broke the ice in China’s large model sector, playing a crucial role in revitalizing the entire industry.
03. Zhu Xiaohu’s Confrontation with Kimi’s Equity Structure
In February, the Dark Side of the Moon announced the completion of a funding round exceeding $1 billion. In November, five investors of Circular Intelligence initiated arbitration over the equity issues of the company.
This event serves as a cautionary tale for equity structures, capital division, and management processes.
04. ByteDance Fires the First Shot in Large Model Price Reduction
On May 15, ByteDance’s AI model, DouBao, announced a price cut, with the market pricing for mainstream models set at 0.0008 yuan per 1,000 Tokens. Subsequently, several manufacturers followed suit with price reductions.
ByteDance’s sudden price cut was a strategic surprise, sounding the alarm for other players.
05. Confusion Over Basic Math: 9.11 vs. 9.9
Mid-year, numerous large models stumbled over the simple math question of which is larger, 9.11 or 9.9.
This event vividly highlighted the ‘intelligence deficit’ of large models, reminding us to exercise caution when applying them.
06. Hardware Vendors Launch AI Laptops and AI Phones
In April, Lenovo launched the AI PC. 2024 was dubbed the “Year of the AI Phone,” with OPPO, Honor, and others releasing concept products.
Hardware, as the physical carrier closest to consumers, is most likely to break through in application scenarios.
07. Tencent’s Hunyuan Quietly Opens Source
Tencent launched the new generation of large model “Hunyuan Turbo.” On December 3, Tencent’s Hunyuan model officially launched video generation capabilities and open-sourced the video generation model.
Although Tencent’s Hunyuan model lags in the base model, it was the first to be open-sourced, making it a significant AI event globally.
08. Baidu’s Wenxin Yanyan Renamed to Wenxiaoyan
In June, Baidu released Wenxin 4.0 Turbo. In September, Baidu announced that “Wenxin Yanyan” had been upgraded to version 4.0 and renamed “Wenxiaoyan.”
Baidu maintains its competitiveness in the large model field but needs to address issues such as business models.
09. iFLYTEK’s Xinghuo Spreads Wildfire
iFLYTEK achieved several milestones in 2024. In October, Xinghuo 4.0 Turbo was released, with its seven capabilities surpassing GPT-4 Turbo.
iFLYTEK’s exploration in vertical industries exceeds many frontline manufacturers, with its hard power still being underestimated.
10. Kunlun’s All-in Bet on AGI and AIGC
Kunlun released the 400 billion-parameter large language model “Tiangong 3.0” and explored layouts in multiple vertical niches.
Kunlun’s layout in the AI field is relatively diverse, with potential victories in music, gaming, and other areas.
In conclusion, we may have overestimated the value of large models in the short term, but we still severely underestimate the value of AI in the long run.
