Will the AI Algorithm Market Boom in 2020?
-
With just over a month left in 2019, a flood of year-end summaries is about to arrive. Looking back at the development of AI this year, perhaps an eight-word summary fits perfectly: 'Declining hype, rising substance.' Many may feel that discussions about AI have cooled down around them. Indeed, this year has seen a significant drop in the number of AI-related investments and mergers, media discussions, and social media buzz compared to the previous two years.
However, if we shift our focus to the actual AI development and application industries, we’ll find that this year’s progress in AI has been remarkable. Innovations in foundational algorithms, frameworks, chips, and the evolution of supporting IT industries in China’s AI sector have made 2019 a fruitful year. If we had to point to one thing proving that AI is still hot, it would be this: AI algorithm markets have begun to emerge one after another this year.
The concept of an algorithm market is easy to understand—it’s similar to an app store. Developers upload their trained AI algorithm models to a shared platform, and users can download them as needed for a fee.
Yet, today, the algorithm market model hasn’t gained much traction, and few outside the AI field even know it exists. Why is that? Why are major AI giants optimistic about this nascent market? Can this new business model take off next year? Let’s delve into the story of the 'spark' that is the algorithm market.
It’s not hard for anyone with a basic understanding of AI to conclude that algorithm markets must be built and will eventually thrive. This is because such a business model aligns perfectly with the logic of AI-to-B development—it’s almost an inevitable choice. Why?
We know that algorithm models are the final manifestation of AI, essentially the software form of deep learning. When AI algorithms start being sold as commercial products, the market will inevitably see fewer algorithms aiming to dominate the most universal scenarios.
This can be called Algorithm Economy 1.0, the early stage of the industry. The beneficiaries of this phase are machine vision algorithm companies like SenseTime and Megvii, with facial recognition and smart security being the most representative applications.
As AI continues to evolve, computing power and data will be deeply utilized, and algorithms themselves will improve in accuracy and broaden in application. At this stage, AI’s capabilities will no longer be limited to simple recognition and judgment but will extend into various vertical needs.
Today, AI can be applied in scenarios we might not even imagine. In smart security alone, capabilities have already been refined to include hardhat detection, drug user identification, and street pet recognition. Moving forward, no two industries or companies will use the same AI algorithms, and single enterprises will employ numerous algorithms in combinations to create their own capability clusters. In this context, neither transitioning unicorns nor AI giants like Huawei and BAT can train all algorithm models in-house. Thus, this phase must introduce a developer ecosystem, similar to mobile internet, where large companies transition into infrastructure and platform services, leaving vertical algorithm development to AI developers. This creates an ideal model of countless developers + a single platform + countless enterprise users.
This cycle can be termed Algorithm Economy 2.0—and the emergence of algorithm markets will be the defining feature of its onset.
From another perspective, the rise and development of algorithm markets can serve as a 'barometer' for whether AI is being widely adopted by enterprises. Moreover, the model of algorithm markets + numerous AI developers is widely seen as draining the红利 of Algorithm Economy 1.0. As highly generalizable AI scenarios are quickly saturated, the ensuing vertical markets won’t sustain massive unicorns, leaving room for small, agile developer teams to thrive long-term. This also partly explains why AI unicorns have focused on转型 and going public this year. Moving into hardware and platforms and rushing to IPO can be seen as a warning to the broader developer community.
How far off the绕过 of unicorns remains unclear, but algorithm markets are already here.
In the past, when tech giants entered an industry, it would immediately attract capital and media attention. But today, with industrial intelligence and industrial internet as the main themes, giants are focusing more on platform infrastructure and technical hotspots, which rarely draw external interest. Algorithm markets fall into this category. In fact, 2019 can be called the inaugural year of AI markets, with AI algorithm model markets 'springing up' and showing notable activity.
As early as March, at Huawei’s China Ecosystem Partner Conference, Huawei Cloud launched its AI Market, featuring modules like AI Model Market, API Market, WIKI Dataset, Competition Hub, and Case Hub. The AI Model Market allows developers and users to publish and subscribe to AI algorithm models.
Soon after, Alibaba Cloud released version 3.0 of its machine learning platform PAI, introducing an algorithm model market segmented into dozens of scenario-specific zones.
In May, Baidu upgraded its AI Market at the Baidu Brain Ecosystem Partnership and AI Market launch event. Baidu AI Market 2.0 offers more diverse展示 and sales options, emphasizing buyers’ ability to post需求 for customized services from AI developers.
By year-end, algorithm markets in various verticals began to emerge, and resources nurturing the AI development ecosystem grew. It’s reasonable to believe that soon, the competition will shift to attracting top buyers and sellers, as优秀 developers and users are稀缺 resources in the AI产业链. In a many-to-many algorithm market model, a single high-quality供需点 can drive significant交易 activity. Categorizing today’s algorithm markets reveals three primary商业目的支撑 their creation.
Regardless, 2019至少 marked the first phase of AI market emergence. The next question is obvious: Will algorithm markets boom next year?
While the algorithm market’s business model is a logical inevitability in the long term, like AI-to-B itself, it requires iterative progress between development and application ecosystems. For example, browsing some algorithm markets today, you’ll find that供需依然 revolve around basic machine vision applications, differing only in识别 objects or pricing. In this context, while爆发 is possible, the更大概率 is a等待 period.
The current landscape shows that while AI developers are increasing, the行业价值,应用潜力, and方案复杂度 of their models仍需提升. Moreover, buyers are明显不足, making算法经济 a challenging primary收入来源 for small开发者 and startups. To change this, algorithm markets and developers may need to explore deeper economic possibilities through these avenues:
Broadly, today’s AI development shows that developers’热情 and能量 are abundant. To prevent过剩, sufficient产业空间 is needed to释放 this energy. Only when AI’s使用深度 and user base reach certain standards can the尴尬 of a火热开发圈 but lukewarm market fade.
A recent hot topic is that as人口红利 fades, China must转向技术红利 and工程红利. With产业智能逻辑 solidifying, AI infrastructure improving, and产业市场 acceptance growing, AI开发红利 may be the first to benefit.
AI algorithm markets are a necessary but insufficient condition for产业智能兴起. In other words, a thriving algorithm market doesn’t guarantee success, but its absence certainly hinders产业兴旺.
In 2020, AI continues to explore while waiting and wait while exploring.