Digital Intelligence 2024: The competitive landscape of large-scale models is established, breakthroughs are made in brain-computer interface technology, and the cloud computing and big data service is transforming | Year-end Review of Workplace Benefits ④
"Workplace Bonus" (ID: WorkplaceBenefit)
Landing and growth are the common wishes of digital intelligent innovation enterprises this year.
The popularity of artificial intelligence remains unabated. The ability of large models has shifted from conversation to multimodality, and the base players in the field have basically been determined. The talent demand in the recruitment end for basic research and development and underlying training has stabilized. Currently, the focus is more on the segmentation of large models and the research and development of product-side landing. The demand for AI product managers and operational positions has increased, and there are more specific portraits and requirements for the project experience of AI talents. The reaction of the investment market shows that embodied intelligence is becoming the next direction that everyone is optimistic about.
AI has also brought a growth opportunity to big data and cloud computing. AI makes data processing easier, and the demand for professional data scientists in enterprises is decreasing. Instead, the demand for basic data engineers, machine learning engineers who optimize models, and data product managers is increasing; cloud computing service providers have launched enterprise-level AI cloud platforms, and the corresponding demand for cloud platform planning, deployment, development, service, and operation and maintenance positions has also increased.
Beyond AIGC, brain-computer interfaces have shown new vitality. This year, Musk made brain-computer interfaces popular, Mihoyo invested in a brain-computer interface company, and brain-computer interfaces have been frequently seen. Policy support and industrial layout have also brought new development opportunities to the industry.
Generative AI: Base players determined, more specific talent demand at the recruitment end ╱ 01
Cloud Computing and Big Data: Serving AI, the gap in cloud computing positions is narrowing ╱ 02
Brain-Computer Interface: The industry sees the dawn, and implant technology becomes the key direction ╱ 03
Generative AI: Base players determined, more specific talent demand at the recruitment end
With the further maturity of generative AI, the talent demand for large models in basic research and development and underlying training has stabilized. According to the 2024 Large Model Talent Report released by Maimai, The talent supply and demand ratio in the large model field is 1.76, with an overall oversupply.
After the Hundred Model Battle in 2023, the large models have basically competed out the top "Six Tigers": The Dark Side of the Moon, MiniMax, Zhipu Huazhang, BaiChuan Intelligence, Stepwise Star, Zero One Everything, and the "Extra Tiger" DeepSeek. At the same time, TMT companies such as Baidu, ByteDance, Alibaba, 360, and Zuoyebang have also emerged.
Among the remaining large model factories, the To-C strategy of The Dark Side of the Moon seems to be more sophisticated.
Kimi raised the banner of "Long Text" as soon as it was launched, aiming at the usage needs of workers, and took the lead in attracting a large wave of traffic. MUA (Monthly Active Users) has long been in the top 3 of the AI product list. The rising star Doubao, thanks to the precise distribution ability of Douyin, its ability to attract new users is far ahead of other APPs in a cliff-like manner.
"Workplace Bonus" (ID: WorkplaceBenefit)
However, generative AI is no longer limited to text-to-text and text-to-image, and has made breakthroughs in more difficult aspects such as text-to-video and text-to-audio. Keling and Ruisheng have quickly become popular.
"Sitting up in astonishment while on the verge of death" is probably the true portrayal of many large model companies and people concerned about the AI field during the 2024 Spring Festival. After waking up from the New Year's Eve hangover, not only a new day awaits, but also the overwhelming news of the text-to-video Sora in the WeChat Moments. (Note: "朋友圈" is a Chinese social media feature. It might be better to translate it as "WeChat Moments" if the context is related to the Chinese social media platform WeChat. However, without further context, it's difficult to determine the exact translation. For now, I'll keep it as "WeChat Moments" in the translation.)
Many large model companies are intensively laying out the text-to-video large model: Kuaishou launched Keling, ByteDance launched Jimeng, and Tencent launched Hunyuan Text-to-Video. In order to recruit talents related to the text-to-video large model for research and development, a recruitment boom for related positions has emerged in China this March, involving positions including Multimodal Large Model Algorithm Engineer, Front-end and Back-end Development Engineer, Deep Learning Framework R & D Engineer, and Pre-training. Among them, the average monthly salary of Digital Front-end Engineer is 67,000 yuan, ranking first in high-paying positions.
Kuaishou launched Keling with the same architecture as Sora in June. Keling has also lived up to expectations and performed well in the text-to-video field. According to the analysis tool Similarweb, as of September 24, the total visits of Keling reached 33.7 million, higher than Runway (31.34 million) and Pika (7.52 million), which had already released products.
While OpenAI, which announced the text-to-video large model at the beginning of the year, only officially released Sora in December. After the release of Sora, an internal message from Kuaishou leaked out: Kuaishou attaches great importance to Keling. They are not recruiting much for other positions, but there are still many HC for Keling. According to 36Kr's "Intelligent Emergence", ByteDance also raised the priority of Jianying Jimeng [1] during the same period.
"Workplace Bonus" (ID: WorkplaceBenefit)
● Kuaishou Keling AI Special Recruitment Poster
Landing applications is the consensus of large model enterprises this year, and there are already 15 AI innovation tracks in the application end testing the waters [2]. Among them, MiniMax's Talkie overseas is the most successful. According to the estimation of 36Kr's "Intelligent Emergence", MiniMax's revenue in 2024 is likely to exceed 70 million US dollars. In September 2024, OpenAI released the o1 inference model, which made Slow Thinking popular [3]. While the large model manufacturers are landing applications and continuously updating the reasoning ability, the talent demand has also changed significantly compared to 2023.
Workplace Bonus learned from the headhunters concerned about the AI field that this year, large model enterprises are more focused on the segmentation of large models and the research and development of product-side landing. The demand for AI product managers and operational positions has increased, but there are more specific portraits and requirements for the project experience of AI talents; High-end technical positions, such as cloud computing and algorithm positions, are still in short supply.
"Workplace Bonus" (ID: WorkplaceBenefit)
In addition, the salary increase of large model positions remains at 30%-50%, and Beijing has become the main gathering place for large model talents. Among the enterprise recruitment needs, the Internet industry ranks first, and ByteDance and Xiaohongshu are the major recruiters of large model positions.
This year, generative AI has also achieved end-to-end, giving self-driving a more possibilities. Tesla, Baidu, Yuanyong Qixing, Xiaoma Zhixing and other companies have made breakthroughs in the end-to-end self-driving field.
In July 2024, Baidu's driverless taxi, Luobo Kuaipei, was piloted in Wuhan, causing a "collective protest" from netizens. People believe that the development of artificial intelligence is bound to lead to the loss of real people's jobs, but taxi drivers in Wuhan do not think so. It is true that the driverless technology is still in the early and middle stages of development, and it is too early to discuss "Replacing real people". At the same time, the recruitment end has a high demand for research and development, especially in the fields of Battery Technology, Self-driving Algorithms and Vehicle Electronic Architecture, and the salary of positions has also increased, with an increase of 20%-30% in the past year.
In May 2024, Meta's initial success also allowed big companies to see the future of the AI hardware consumer market. Meta launched smart glasses based on generative AI. According to The Verge statistics, as of May 2024, the global sales of this smart glasses have exceeded 1 million pairs - No previous smart glasses have had such high sales.
Domestic AI hardware is also following closely. ByteDance launched AI headphones, and Baidu launched AI glasses to occupy the track. In addition, brands that are interested in laying out the AI glasses project and are evaluating include: OPPO, vivo, Huawei, Tencent, ByteDance. According to the announced news, ByteDance angel-round invested in Li Weike, founded by the former product manager of Tmall Genie, mainly doing AI glasses; Xiaomi and Goertek cooperated to lay out AI glasses, which are expected to be released in Q2 2025.
For job seekers, the rise of the AI hardware trend will also lead to an increase in the demand for corresponding Hardware Product Managers, Architects, and Dispatchers. However, the C-end feedback of Baidu and ByteDance's two products is not good. The seemingly emerging AI hardware is still in its infancy. In this regard, for Job seekers whose original main business is not related to AI hardware, it is recommended to remain vigilant at all times.
In the past year, large model enterprises have not found an explosive scene in the To C trial. The endgame of generative artificial intelligence has turned to accelerating scientific and engineering capabilities. Many big shots such as Yu Yinan, VP of Horizon, Chen Yilun of Huawei Vehicle BU, Gao Jiyang, former mass production R & D head of Momenta, and Xiao Zhiguang, the 10th employee of Xiaopeng Motors, have started embodied intelligence entrepreneurship.
The embodied intelligence financing market is booming and has become the next direction favored by capital. An investor concerned about the direction of artificial intelligence told Workplace Bonus that The investment market is also looking at some embodied intelligence companies now.
[1] Refer to the article in 36Kr's "Intelligent Emergence". The article mentioned that ByteDance's management judged that AI conversation-based (or chatbot-based) products may only be an "intermediate state" of AI products, and the more ideal product form in the long term is likely to require a more visual user experience and a lower user usage threshold. Therefore, ByteDance has raised the product priority of Jimeng and is trying to create the "TikTok" of the AI era with a new path.
[2] Refer to the sharing of Cai Suqing, the founder of 36C at the 36Kr WIES Conference 2024.
[3] Slow Thinking is a thinking mode that requires active control and conscious thinking. It relies on rationality and logic and is suitable for dealing with complex and novel problems.
"Workplace Bonus" (ID: WorkplaceBenefit)
Cloud Computing and Big Data: Serving AI, the gap in cloud computing positions is narrowing
The B-side of the obvious fatigue of the Internet is the slowdown in the growth rate of cloud computing and big data service providers. However, the arrival of AI has injected new vitality into the industry.
The demand for computing power from large model enterprises is continuously rising. In May, computing power cloud service providers have all lowered prices to start a price war, using low prices to attract AI-related enterprises to settle on the platform. Since the entire cloud service market seems to have no new customer demand, cloud service giants are also investing in large model unicorn enterprises, hoping that the latter can consume more cloud computing power. For example, Alibaba has invested in 5 large model enterprises.
"Workplace Bonus" (ID: WorkplaceBenefit)
Large models have indeed brought considerable growth to cloud service providers. Taking the data of the leading manufacturer Alibaba Cloud as an example, from the fiscal year 2019 to the fiscal year 2023, the revenue growth rates of Alibaba Cloud are 84%, 62%, 50%, 23%, and 3.5% respectively, showing a decreasing trend year by year. According to the financial report disclosed by Alibaba Cloud in the third quarter of 2024, Alibaba Cloud's revenue increased by 7% year-on-year to 29.61 billion yuan, and this is also the fourth consecutive quarter that Alibaba Cloud has achieved a growth rate increase.
AI has also become the focus of