Has Artificial Intelligence Cooled Down?
-
In 2019, many claimed that artificial intelligence (AI) had lost its momentum. But is this really the case? The author analyzes the situation from the perspectives of talent supply and demand, sharing their insights and conclusions.
Graduating in 2017, inspired by the book On the Top of the Waves, I resolutely chose to ride the wave of AI, hoping to seize life-changing opportunities. Over the past three years, I’ve grown into a different version of myself, grateful for that decision. From a clueless graduate, I’ve evolved into a product manager familiar with dialogue systems, algorithm principles, and product design, with a clearer career path ahead.
2019 was notably cold for the internet industry, with layoffs at major companies and startups shutting down. Many AI professionals around me expressed pessimism, feeling that the reality of AI fell far short of their initial expectations. A recent question on Zhihu prompted me to reflect on the industry: Is today’s AI industry the same as it was three years ago? Has its direction shifted? Do I need to adjust my career path?
Here’s my conclusion: The AI industry remains vibrant. Despite economic downturns, businesses are eager to cut costs and improve efficiency, sustaining strong demand for AI. The perceived cooling of the industry and job scarcity stem from a mismatch between talent supply and demand.
Compared to mature sectors like retail and finance, AI is still a niche field. It struggles to create standalone value, instead relying on transforming traditional industries—reducing operational costs and uncovering new growth opportunities. This reliance means AI doesn’t require the massive workforce typical of traditional industries.
Yet, the AI boom has attracted a flood of entrants: graduates from new AI programs, non-AI majors, and career switchers via training institutes. As AI penetrates deeper into industries, the 'low-hanging fruit' of cost-saving and efficiency gains diminishes, raising the bar for talent. Beyond basic AI skills, industry-specific expertise becomes critical. Many AI practitioners lack vertical experience, leading to mismatches with employer needs.
Years of development have produced a pool of experienced AI professionals. Universities, aligning with policy trends, have launched AI programs, injecting fresh talent into the market. Training institutes also contribute hybrid professionals. Increased supply intensifies competition—though healthy competition drives progress.
Early in AI’s rise, most AI product managers transitioned from traditional roles or doubled as algorithm engineers. These pioneers, deeply versed in business or technology, gained invaluable 0-to-1 project experience, making them highly sought-after.
Initially, AI projects demanded vast data labeling, often handled by internal teams. Post-launch, AI trainers took on operational roles, analyzing logs and refining systems, eventually transitioning to product management. Recent AI graduates also join as product managers, though they need time to build experience.
AI trainers are now polarized: those limited to basic labeling face declining wages, while those skilled in model principles, bug fixes, and workflow design command premium salaries. The broad label of 'AI trainer' inflates hiring costs. Smart upgrades in customer service and crowdsourced platforms have flooded the market with labelers and quality checkers.
In customer service, labelers often come from frontline roles. They excel at knowledge base organization and data tagging but lack operational and technical depth, hindering post-launch improvements. Basic trainers earn ¥3.5k–8k/month, while team leaders fetch ¥12k–16k.
Algorithm engineers hail from AI programs or self-taught backgrounds. While media hype touts million-dollar salaries, only a few elite engineers—with robust project experience and optimization skills—reach such heights. Bootcamp graduates vary widely in skill, often relying on open-source solutions without deep algorithmic or industry-specific modeling expertise.
Tech giants like Baidu, Alibaba, and Meituan embraced AI early, integrating it with their operations and refining it over time. For instance, Meituan evolved from machine learning to deep learning for delivery time estimates, ensuring high availability through lightweight optimizations.
Today, user-facing departments at top firms rarely hire AI roles. When they do, they seek product managers with years of vertical experience, requiring only basic AI knowledge. They aim to deepen applications and unlock more value. Algorithm engineers must model scenarios and solve business problems.
Having polished internal products, these firms now export solutions (e.g., Alibaba Cloud, Baidu AI Open Platform). Their B2B divisions prioritize AI professionals with project experience, communication skills, and solution design—delivering standardized products with tailored implementations.
Second-tier firms vary widely: some embrace AI; others adopt cautiously or reject it outright.
Companies that take a rational approach to AI typically procure targeted AI systems from suppliers to address specific business scenarios, avoiding the need to build multiple machine learning environments from scratch, accumulate and label data, and bear the costs of maintaining an entire AI project team.
Companies that actively embrace AI generally prefer candidates with experience in building systems from the ground up (0-1 experience) to help them quickly establish their infrastructure. If a project is imminent and no candidates with such experience are available, they may settle for those with some AI experience.
The rise of any technological wave inevitably attracts pioneers who establish startups. In the AI industry, early startups include consumer-facing ventures like Mobvoi (a personal assistant service) and B2B-focused companies like Zhuiyi, Zhujian, and AISpeech.
B2B enterprises typically follow a "usable-standardized-scalable" business model. However, the AI industry currently struggles to independently generate commercial value. Against the backdrop of an internet winter, AI startups often collapse more quickly. Consequently, they usually maintain a small innovation team alongside multiple project implementation teams.
Startups, finding it difficult to attract top-tier talent, often adopt a 1+1 approach to achieve their goals: hiring someone with complementary skills to pair with internal staff. For example, when Startup A aims to develop a smart government solution, it might hire someone with government experience and pair them with an internal AI product manager to jointly complete the solution's development and deployment.
Every industry evolves from inception to development to maturity. Mature industries are either dominated by a single player or engaged in cutthroat competition among a few evenly matched companies, with none gaining significant advantage. Many mature companies seek AI transformation to develop new business lines and achieve differentiated competition.
These companies often prefer candidates with 0-1 experience to help them quickly establish their systems.
The trend of enterprise development is irreversible and continuously deepening. Therefore, aligning one's skills with corporate needs and growing alongside the company is crucial. How to stay at the forefront of this wave will be the topic of the next article, which will outline the skill set required for AI practitioners.