The rise and proliferation of AI is truly a sight to behold. Whether you're a part of the crowd that's embracing the tech and all its wonders, or the group that fears it's Skynet reincarnated, you have to give props for its transformative power on a global scale.
Yet, for all its wonders, the current centralized approach is problematic, to say the least. Not only is the majority of the AI pie held tightly by a few private companies (that are still the de facto leaders of the market), but there's also the so-called black box problem. This is an intrinsic flaw of centralized AI systems, where it's impossible to tell how AI makes its decisions.
There are more and more decentralized projects looking to help break up the monopoly that private companies have over AI, and potentially solve the centralized AI black box dilemma. Among these is OORT, a data cloud for decentralized AI.
The team behind OORT quickly realized that the hardware required to develop new models costs an arm and two whole legs, leaving only the key players like Google and OpenAI with the funds and the juice to support the rapid development of the technology.
To address the hardware and compute issue, OORT created one of the few fully-functional Web3 AI ecosystems, where the community can pitch in their computational resources to provide decentralized computing power and enable data handling not hampered by geography. This is achieved with "Deimos II" Devices, which serve as personal edge nodes.
The first incarnation of "Deimos I" logged over 51k devices scattered across 22 countries. This success led to the launch of "Deimos II," which implements on-device LLM interference that allows users to run small AI models locally without sending data to the cloud.
However, hardware and computation are just a part of the AI puzzle, as founder and CEO, Dr. Max Li, tells Tech Times.
"When we talk about AI's development, it's always three key pillars: data, algorithms (models), and computing power. For many companies, data is the component they cannot easily acquire. This convinced me that in the future, data will be the only bottleneck."
Glass Box vs No Box?
All AI systems developed by private tech giants are running on centralized infrastructures and are trained on proprietary datasets. It doesn't take a genius to conclude that the end user has zero insight into the inner workings of these machines, making it super difficult to trust the results.
The datasets may not be inclusive and representative, potentially leading to bias. It's also impossible to audit the training process. This leaves all of us operating in a landscape where most are barred from truly innovating, and it's hard to have faith in mysterious AI models. The status quo never changes.
One potential solution for the black box issue is turning the black into a "glass." Referred to as glass box modelling, this approach requires training data that can be explained, modified, and examined by analysts. The end result would be an AI application that mimics human-based decisions, guaranteeing that the algorithm itself has been tested for accuracy and can ultimately be explained.
While this is better than nothing, it's realistically just another limitation we'd have to deal with—one that would still require us to trust the company that owns the infrastructure and the data. Basically, we'd be back to square one.
OORT aims to remove any boxes whatsoever by introducing DataHub. This is an integral part of the decentralized ecosystem that allows the community to contribute and annotate decentralized data securely. Combined with OORT's distributed Storage & Compute network, it represents one small step out of the black box.
"While the chip war focused on producing the most powerful hardware, the data war hinges on acquiring the right datasets to train AI," Dr Li states.
"The growing scarcity of ethical, high-quality data poses a dilemma for businesses of all sizes, from giants like Google and Microsoft to smaller companies struggling to access training datasets. For large companies, acquiring data from centralized giants may still be feasible, albeit costly," Li further clarifies.
"However, smaller businesses face limited and often unaffordable options. Without access to proper methods or channels for data collection, these companies risk being left behind in the race to innovate."
Considering that OORT's decentralized cloud incentivizes a broader community to contribute and participate, it's a viable solution that, in theory (and if it reaches mainstream acceptance), could solve the black box issue and cut the reliance on privately owned companies.
"The blockchain allows everyone to contribute organic data from anywhere in the world while also receiving incentives," says Li. "This is further supported by blockchain's ability to facilitate small-amount, cross-border transactions within minutes, a capability that makes such a model nearly unthinkable for traditional companies reliant on conventional banking payments."
OORT is confident that the future of AI is fully democratic, claiming that the success of community-sourced OORT datasets (which reached the number one spot in multiple categories on Kaggle, Google's largest data science community) only shows that many people are gravitating toward a more fair approach, without maybe even being aware of the black box problem.
Granted, it may take a few years for individuals and institutions to break up the monopoly of the few tech titans, but there's no denying that AI will ultimately belong to the people, free of any boxes or limitations.
ⓒ 2025 TECHTIMES.com All rights reserved. Do not reproduce without permission.
