Researchers have created an innovative framework designed to make complex AI tasks easier and clearer for users to understand. This framework focuses on giving solutions for deep reinforcement learning (DRL) inquiries. It connects developers, companies, and individuals with specific AI needs to service providers equipped with the necessary resources, expertise, and models. The service operates on a crowdsourced basis, utilizing blockchain technology and smart contracts—code-embedded agreements with predefined conditions—to link users with suitable service providers.
In the future, workplaces will operate on vast amounts of data. To effectively handle this, businesses, developers, and individuals will require enhanced artificial intelligence (AI) systems, better-trained AI personnel, and more powerful data processing servers.
While major tech companies have the capability to fulfill these requirements, many small and medium enterprises and individuals find these resources out of reach. To address this gap, a team of international researchers from Concordia University has created a new framework that aims to make intricate AI tasks more attainable and straightforward for users.
Outlined in a paper published in the journal Information Sciences, this framework focuses on offering solutions for deep reinforcement learning (DRL) requests. DRL, a branch of machine learning, merges deep learning—which employs layered neural networks to identify patterns in extensive datasets—with reinforcement learning, where an agent learns to make decisions through interaction with its environment based on a reward and penalty system.
DRL finds applications in various fields including gaming, robotics, healthcare, and finance.
This new framework connects developers, companies, and individuals with niche AI requirements to service providers who possess the needed resources, expertise, and models. Built on a crowdsourced platform and blockchain technology, it employs a smart contract system to ensure effective user-service provider matching.
“By crowdsourcing the training and design of DRL, we enhance transparency and accessibility,” states Ahmed Alagha, a PhD student at the Gina Cody School of Engineering and Computer Science and the lead author of the paper.
“With this framework, anyone can register and build a history and profile. Based on their skills, training, and ratings, they can take on tasks requested by users.”
Making DRL Accessible to All
Co-author Jamal Bentahar, a professor at the Concordia Institute for Information Systems Engineering, emphasizes that this service will broaden the reach of DRL to a larger population than ever before.
“Training a DRL model requires computational resources that aren’t accessible to everyone, along with a certain level of expertise. This framework provides both,” he explains.
The researchers assert that their system’s framework will lessen costs and risks by spreading computational tasks through the blockchain. The potentially severe effects of system failures or cyberattacks can be diminished by utilizing numerous machines addressing the same task concurrently.
“If a centralized server goes down, the entire platform collapses,” Alagha points out. “Blockchain offers distribution and transparency. Every action is recorded, making tampering exceedingly difficult.”
The challenging and expensive process of training a model can be expedited by utilizing an existing model that requires only minor tweaks to cater to specific user needs.
“For example, if a large city creates a model to automate traffic lights for better traffic management and fewer accidents, smaller towns may not be able to develop a similar model independently. They can, however, adapt the existing model crafted by the larger city for their own situations.”
Contributors to this study include Hadi Otrok, Shakti Singh, and Rabeb Mizouni from Khalifa University in Abu Dhabi.
