Recently, I've been pondering why the field of artificial intelligence (AI) and machine learning (ML) has been so successful, achieving significant breakthroughs continuously. I prefer analyzing problems from a systemic perspective, so I took some time this weekend to write down my thoughts.
I believe that the AI field has optimized many key aspects of its ecosystem's internal processing algorithms, allowing for the rapid creation of value, discovery of talent and technology, publication of findings, and attraction of capital. The ultimate test, of course, is its benefit to humanity. Here are some critical components of this ecosystem I've identified.
A crucial element is the presence of fair and open leaderboards. Despite my modest involvement in research and development, I've noticed many so-called experts select advantageous testing or validation methods to claim their solutions are the best. This approach may make every paper seem excellent, but it fails to identify truly outstanding work, often resulting in no overall progress.
The AI field benefits from numerous open-source, relatively fair, and widely recognized leaderboards. Open-source datasets or testing methods and quantitative indicators are used to verify and test whether an AI model, algorithm, or engineering improvement genuinely advances technology.
These fair leaderboards allow researchers to quickly identify the best algorithmic tools. Users can find the best tools for their unique problems through these leaderboards. Meanwhile, investors and sponsors can see which teams are performing best, regardless of their country of origin, gender, or other factors. If you excel, significant funding will support you. This is the first key node.
Another critical aspect is the publication of papers. In AI and ML, new findings are first released as preprints on free websites like https://arxiv.org/. This approach ensures the swift sharing of research advancements, achieving zero-delay dissemination, allowing the world to access findings without cost. This zero-delay, zero-cost distribution of new technology is vital.
In other fields, publishing and disseminating work face significant barriers. Papers, regardless of quality, sometimes need multiple submissions to various conferences. If lucky, a good review gets it through; if not, it might get buried without further effort. Thus, the publication process in some fields can drag on for years, significantly delaying the spread of good technology. Additionally, because reviewers act as intermediaries, they often bury good work.
Furthermore, the cost-free accessibility of papers to the world is another hurdle in many fields. Often, papers are behind paywalls, requiring payment to access, which is a poor practice. Many research teams rely on public funding, yet publishing requires payment to publishers. Therefore, AI's ability to offer cost-free, barrier-free access to research findings is a significant advantage.
Moreover, the AI and ML fields excel in open sourcing-, not just with code but also with many models and datasets. With abundant open-source software, models, and datasets, other researchers can easily build upon the giants' shoulders, making further improvements, proposing new algorithms, and new engineering methods, thus effectively advancing the R&D process. This open-source spirit in AI is highly commendable.
Another point is the almost real-time, 24/7 global collaboration and competition, uniting the world's smartest individuals in an incredibly engaging game.
Thanks to internet development, the entire process from initial idea to implementation through software and data, and finally to paper publication, has been significantly optimized. This optimization allows global researchers to quickly catch up on the latest progress, easily find collaboration partners, and together create new things. This cross-border, -political, -religious, and -cultural collaboration, driven by a shared goal of advancing technology for humanity's benefit, is a key reason for AI's significant progress.
Overall, from a systemic and ecological perspective, any research and development field, not just AI or ML, requires a well-functioning ecosystem to thrive. The essence is a carefully designed and optimized algorithm for the internal coordination of various system elements. An optimized ecosystem can advance each step quickly and cost-effectively, ultimately creating value for society and providing substantial returns to researchers.
For individuals in research and development, we should aim to join or improve the ecosystems of our fields. This can include creating and maintaining fair and open leaderboards, encouraging the public release of data and software, and allowing papers to be published with minimal delay on platforms like arXiv. If your ecosystem is lacking, it's like being on a sinking ship; individual efforts can hardly offset the ship's decline or even its sinking.