Recently, I have been thinking about why the research and development field of artificial intelligence (AI) or machine learning (ML) has been so successful and continuously achieves significant breakthroughs. I like to consider problems from a systems perspective, so I spent some time analyzing this over the weekend. I believe the AI field has optimized many critical ecosystem processes and algorithms, enabling it to rapidly create value, identify exceptional talent, advance technology, publish achievements, and attract capital. Ultimately, the gold standard for evaluating success is the benefit to humanity. Below, I summarize several key nodes of this ecosystem.

Fair and Transparent Leaderboards

Firstly, AI benefits significantly from having fair and openly accessible leaderboards. This is critical because, even in my humble experience in research, I’ve noticed many so-called experts tend to select testing methods favorable to themselves when writing papers or marketing products. As a result, every paper looks excellent, but the downside is that truly outstanding innovations become hard to identify. Often, efforts are fragmented rather than genuinely progressive.

The AI field offers numerous open-source, relatively fair, and widely recognized leaderboards. These allow researchers to verify their AI models or algorithms comprehensively using open datasets, standardized testing methods, and quantifiable metrics, genuinely advancing technological progress.

With these unbiased leaderboards, researchers can quickly and clearly see which algorithmic tools are currently best. Users can swiftly identify optimal tools for their unique challenges. Meanwhile, investors and research sponsors can easily identify top-performing teams, irrespective of nationality, gender, or other factors. If your work ranks highly, ample funding awaits. This is the first critical node.

Rapid Publication and Dissemination of Papers

Another crucial factor is the rapid publication of papers. In AI and machine learning, when new results or discoveries emerge, researchers first upload their preprints to free websites like https://arxiv.org/. This allows for the immediate sharing of the latest research, achieving zero delay and enabling global access to these results at no cost. This immediate, cost-free dissemination of cutting-edge technology is extremely important.

In many other fields, two significant barriers prevent rapid dissemination. Papers often undergo multiple submissions to various conferences, facing the risk of rejection due to unfavorable peer reviews. Valuable work may thus be buried. Consequently, publishing processes in other disciplines can take one or two years, significantly delaying technological advancements. Additionally, papers in other fields are frequently behind paywalls, costing readers tens of dollars per article, despite research often being funded by taxpayer money.

The AI field’s ability to freely and immediately share research without artificial barriers allows everyone to instantly access the latest developments. This practice is highly commendable.

Open-source Code and Datasets

Moreover, machine learning and AI excel in open-source initiatives, offering not only open-source code but also numerous publicly accessible datasets. With abundant open-source software and datasets, researchers can easily build upon existing advancements, propose new algorithms, and devise innovative engineering methods, effectively propelling research forward. The open-source spirit within the AI community is highly praiseworthy.

Global Competition and Cooperation

Another remarkable aspect is globalized, nearly real-time, 24-hour continuous collaboration and competition. This resembles a fascinating game among the world’s brightest minds.

Due to internet advancements, the entire process from initial idea to software and data implementation, culminating in paper publication, has been extremely optimized and nearly perfected. This allows global researchers to rapidly see the latest developments, easily find collaborators, and create new innovations together. Such cross-border cooperation, transcending political, religious, and cultural differences for the collective benefit of humanity, is a major reason behind the tremendous advancements in AI.

Conclusion

In conclusion, from a systemic and ecological perspective, any research and development area—not just AI or ML—requires an efficiently operating ecosystem for support. Essentially, careful design and optimization of algorithms for coordination among internal ecosystem elements are necessary. An optimized ecosystem quickly and cost-effectively advances each step, ultimately creating societal value and providing substantial rewards to researchers.

For individuals in research and development, joining or actively improving a robust research ecosystem is advisable. Efforts could include promptly creating and maintaining transparent leaderboards, encouraging open publication of data and software, and promoting minimal-delay paper uploads to platforms like arXiv. If your ecosystem is weak, personal efforts alone may not counteract its overall decline.

Note: Translated from https://mp.weixin.qq.com/s/UeXbY_hgUOZ5WhPYoomDAg?token=629299755&lang=zh_CN