With the increasing advancement of AI and ML technologies comes the understanding that our current relationship with AI is deeply shaped by our society’s historical experience with technology. In this course, AI engineers and enthusiasts will gain the skills to design human-centered AI experiences using human-focused science. Learn how to deploy techniques from social science to tackle cognitive gaps and fallacies when building effective computing platforms. Through case studies, practical tips and design anthropology exercises, you observe how anthropology and ML can add a new dimension to AI.

When was the first time I saw Facebook’s new smart glasses “Ray-Ban Stories”? A couple of minutes ago? Ha…No, a couple of years ago, and specifically in Leipziger Pl. 9, Berlin, Germany at the Spy’s museum! In the latter glasses, a tiny digital video camera is integrated into the traditional spy glasses and videos saved into an accompanying recorder. In Facebook’s glasses, it is stored in your smart phones though…

Two main talking points about this product:

Privacy and Adoption:

• Privacy?

Obviously in AI, I do believe that German’s Bauhaus Design Philosophy ” Form Follows Function” does not apply to AI. Even if the LED indicator light flickers on when the Facebook’s glasses are recording, notifying people that they are being photographed or filmed, different types of biases, uncertainties and discriminations will come into surface from the usage of the product.

• Adoption? In a world of deepfakes, we might not be surprised to see a shift in human behaviour that is very similar to the peri-personal space behaviour our ancestors had in the stone age while they experienced stone age technologies. Indeed, we might wear these smart glasses to record interactions with our surroundings so that it protect us from baseless allegations in future automated cities. It also comes with the cool Ray-Ban brand

I speak more about these concepts in my newly released UC Berkeley Extension course here.

I am happy to know your thoughts/comments?

Today’s factories will need to be able to produce customized products with variable requirements, and this will trigger them to become smart factories, which will leverage emerging technologies like AI and the IoT, or what is being called the Industrial IoT (IIoT). In the data economy, industrial companies will need to act fast, since relying on static scheduling in production will not suffice to cope with changing order information. Also, they will need to be able to distribute the workload among robots and machines dynamically, also taking into account the equipment’s energy consumption, which most traditional factories are just not equipped to do.

The Necessary Edge

Personalizing services through dynamic task distribution, in this context, requires a technology that achieves ultra-low latency at the level of the local machines that distribute the production tasks. Because it is a burden to send this data to a cloud manufacturing platform for analysis, edge computing is the recommended approach for this scenario, as it would be able to support real-time production cycles. For a smart factory engaged in high personalization and a variety of machine tasks, predefined rules would not work unless an intelligent edge node is established that can use its own self-organized coordination abilities to negotiate with other nodes to solve largescale problems in parallel. Machines and robots at a factory are now starting to be equipped with edge nodes that can measure energy consumption. These nodes also have analytics embedded, such as optimization algorithms like swarm intelligence, which can prioritize tasks based on each machine’s individual energy consumption.

In the scenario above, each edge node has intelligence and autonomy, as each one is embedded with reasoning, planning, and machine-to-machine coordination abilities. The role of these edge nodes differs from that of a smart meter, which can only estimate the relationship between the workload of one piece of equipment and the energy it is consuming. In contrast, these edge nodes act as data storage, analytics, and intelligent information interaction nodes.

What Does This Edge Platform Look Like?        

This use case, among many others, including predictive maintenance through real-time altering, is not possible without a proper multi-access edge computing (MEC) platform. A MEC platform working with 5G could make it possible to produce ultra-low-latency-sensitive manufacturing applications. Because of the large volumes of data that will be generated by the machines and robots in a smart factory, applications like this will be required, not only to handle the data volumes but also to deliver the needed processing capability in a local context with location awareness. A MEC platform consists of different tiers that collect data from IIoT machines to perform different tasks across each tier, such as reducing redundancy and errors; decoding and compressing the data, as well performing other raw-data pre-processing operations; and storing the data. To be optimally effective, a MEC platform must also be able to support deep-learning-processing-intensive, low-latency features such as real-time anomaly alerts, to meet stakeholders’ needs for enhanced decision-making insights.

Why Does the MEC Platform Need Data Virtualization?

To be optimally effective, a MEC platform requires access to all data sources across all the different tiers. However, MEC platforms that rely on traditional data integration methods like ETL processes will struggle to connect to these disparate data sources. This is why MEC platforms require an agile data integration technology like data virtualization to be fully effective. Data virtualization can be deployed at any intermediary layer between the edge and data center or the cloud, including organization and region-based layers. It works by abstracting the data sources and creating a unified view of all the information, i.e. maintenance, inventory, parts, and dealers, to optimize processes and enable predictive analysis.

Finding the balance between machines and human creativity in a smart factory could be through the emergence of virtual voice assistants that will be the next interface in a smart factory.

• Bill of Materials (BOM) identifies all product parts and the timeframe and cost of production.
• Big Data store, manage, and track BOM or even schedule required materials that should be available to reduce costs.
• 5G could make ultra low latency sensitive manufacturing applications possible.
• Data Virtualization securely combines data from diverse source systems and create unified views that made it easier to take corrective steps to ensure ideal yield and efficient plants.
• Edge computing orchestrate production and task scheduling.
• Virtual voice assistants will be the interface that resonates and talks with humans to track all bill of material goods, assemblies, subassemblies and tracks finished goods and assets in a factory to alert and communicate to us information about their status and condition. These VAs are becoming more anthropomorphistic embodying humanistic traits from voice tonality to design, convenience, sociability, and experience.

Maximize Yield, Minimize Downtime

To keep up with supply and demand, manufacturers can settle for nothing less than highly optimized systems that maximize yield and minimize business downtime. They need advanced technologies that work together and despite this high level of automation and collaboration among machines and robots in a smart factory, humans will always remain central to the manufacturing operations. They will always be monitoring, controlling, and supervising the manufacturing process to deliver new innovations in today’s data economy.

Smart Factories will play a key role in the post corona world. Stay Tuned for the next posts to know more!

* This post was originally published at Datavirtualizationblog.com