It has become clear that many enterprise clients have been creating AI models for specific analytics functions, but without underlying connectivity. Large scale cross-correlation of their findings, particularly in real-time scenarios, is also missing. We have decided to build that large-scale connective model – a single AI instance spanning the enterprise, connecting all outlying models, drawing together their real-time findings, and conducting analytics and pattern analysis across all company operations. It will provide the C-suite with a window into all aspects of corporate operations in real time, autonomously identifying and predicting related challenges, and ultimately providing optimal solutions. The Enterprise Neurosystem will network together all AI instances across a multinational corporation (both in-house and vendor sourced), and tie them into a central intelligence that either makes an autonomous course correction or delivers recommendations to management. And this will all be accomplished in open source. It is the end-state combination of AIOps (network automation, predictive maintenance, logistics and delivery automation, etc.) and Business Intelligence (historical analysis, predictive trend intelligence, etc.) within a single overarching intelligence. It will consist of a unified analytics framework that spans all aspects of corporate operations. Much like the neurology of the human body, it is a series of interconnected AI/ML functions, tailored to permeate and assess every aspect of the business – and to autonomously regulate and optimize all day to day functions. Like the human body, many primary functions will be autonomous, while higher order decisions will be presented for human/machine review. This capability would not only increase efficiencies in every area of network and business operations, but it would enable a corporate operations interface of unparalleled depth and clarity. The process will be fine-tuned by corporate users to reframe the deployment and target additional areas of analysis. They will teach the system to understand these areas of focus and further enable the system to be self-directed. Eventually this AI analytics system will run autonomously by maintaining a communications and analytics web across all aspects of the business. In its end state, the overarching intelligence would draw in and cross-correlate the data from the many areas of company operations, and autonomously provide load balancing for lower level functions – adjusting the corporate “system” in real time. In terms of larger operational and strategic issues, human management would be engaged to create a stronger network effect (human/machine), and deliver the best decision capability. A variety of AI architectures will be reviewed for use in this large-scale architecture, particularly in the area of centralized intelligence. A tiered series of GANs, transformer instances, repurposed gauge CNNs, and even the older UIMA architecture are all under review. Over time, this architecture will be repurposed as an integrated system for humans and their environment, to help manage climate change and assist in the detection of extraterrestrial objects. The academic partners include Stanford SLAC, Harvard Analytics and UC Berkeley’s Data-X program, snd corporate members include personnel from America Movil, Equinix, Ericsson AI, IBM Research, Intel, Kove, Penguin Computing, PerceptiLabs, Seagate and Yahoo!
Session Summary
The Enterprise Neurosystem: The Business Singularity – An Open Source Community
MLconf Online 2021
Bill Wright
Red Hat
Head of AI/ML business development and technical strategy for global verticals
Learn more »
Ryan Coffee
SLAC National Accelerator Laboratory
Sr. Scientist -- LCLS SRD / PULSE
Learn more »