Posted by 
Centralized data repositories are ubiquitous across data-driven industries. These centralized data points act as the brain of decision-making systems, where actionable intelligence is triggered. This centralization of data combined with a hierarchy of authority underpins governments, businesses, and culture. But is it the most effective model when dealing with AI?
SwissCognitive Guest Blogger: Eleanor Wright, COO at TelWAI –”Bring The AI To The Data – Distributed Architecture” Image by storyset on Freepik
This centralized model, however, has its drawbacks. Scaling these models puts a strain on the central node and can lead to reduced response times and slow decision-making. There comes a point where the centralized node becomes overwhelmed by data.
This model of centralized data collection is at the core of many data management and AI systems. Systems such as data lakes consume data from multiple feeds, in multiple formats, and utilize AI and machine learning to extract intelligence, reports, and more. Systems developed by media platforms are also built on centralized architectures, consuming, and redistributing data to facilitate user attention and retention. It is through the centralized analysis of data that real-time intelligence is generated, and long-term statistical data is gathered.
Although such systems enable the analysis of multiple datasets and data fusion, handling such large quantities of data can lead to inaccuracies, bias, and delayed response times. If such systems become saturated with bias, data output will not be representative and could lead to inaccurate decision-making. For example, if suggested YouTube algorithms inaccurately target videos that offend the viewer, their media consumption on the platform will decrease.
Although such bias may not be hugely impactful when consuming content, AI holds the potential to enter every aspect of business and governments, potentially gatekeeping funding and human capital investments. Thus, we must not underestimate the impact of bias and inaccurate decision-making as we introduce more and more AI systems into economic operations.
One potential answer to filtering out bias and delayed response times is employing a distributed architecture. In the same way, governments deploy regional responsibilities to state governing bodies, a distributed architecture that retains activities local to the data and decision-making may improve response times and reduce bias. It is by localizing operations that efficiency and accuracy is facilitated.
As Jacob Williamson-Rea of Carnegie Mellon University states “over the next 5-10 years, AI will evolve from today’s highly structured, controlled, and centralized architecture to a more flexible, adaptive, and distributed network of devices. This transformation, called “AI Fusion,” will bring algorithms to data, instead of vice versa, made possible by algorithmic agility and autonomous data discovery” (Jacob Williamson-Rea, 2021). It is this notion of bringing the algorithms to the data, that will enable smarter more accurate decision-making.
Facilitating such operations in the commercial world, however, will have its challenges. Dependent on networks, processing, data security, and more, facilitating a distributed architecture is no simple task. Such an architecture will require a network base layer to build upon, followed by intelligent data points and algorithms facilitating shared intelligence. Historical data may be issued to data centers and the cloud, facilitating retrospective tools and all this will be made smarter through quantum.
This future of intelligence at the edge and beyond may give rise to the next era of big tech disruption where AI is able to share its intelligence across networks.
About the Author:
Holding a BA in Marketing and an MSc in Business Management, Eleanor Wright has over eleven years of experience working in the surveillance sector across multiple business roles.
Der Beitrag Bring The AI To The Data – Distributed Architecture erschien zuerst auf SwissCognitive, World-Leading AI Network.
