Neal Analytics predictions for 2021: Vision, AI, and the edge

Neal Analytics predictions for 2021: Vision, AI, and the edge

2020 was anything but predictable. The COVID-19 pandemic accelerated digital transformation across industries, launching businesses five years ahead of their roadmap with cloud adoption, remote learning, online shopping, telemedicine, and more.

There was the toilet-paper apocalypse and panic-buying that left store shelves empty in the spring. Zoom became a verb. Sweatpants snuck their way into being a staple of WFH attire. The world went digital.

So, what happens next?

Our experts have a few ideas. Check out the list below for the Neal team’s 2021 predictions.

Vision will be the new frontier in 2021

Similar to the impact the field saw when touch came into play, computer vision will open up new use cases and capabilities. Its impact will be seen across scenarios and across industries.

Development in computer vision will accelerate, taking advantage of edge computing to get closer to the point of data collection and offload AI workloads. In retail and healthcare, vision will play a key role in managing inventory and improving customer experiences, such as out-of-stock detection with StockView.

Vision will also be a highlight in manufacturing and supply chain use cases with applications focused on quality control and assurance.

Deep Reinforcement Learning (DRL) will have a big impact in manufacturing.

Microsoft Project Bonsai released for public preview in 2020, but before that there was not much development moving forward in this field. Now, early adopters in manufacturing are starting to see success from investing in DRL, such as PepsiCo’s solution to produce more perfect Cheetos.

In 2021, we predict more manufacturing companies will try POCs for DRL solutions, and those that see improvements by the end of the year will start to invest in a big way.

In addition to production yield optimization, DRL will have an impact on supply chains as companies rethink operations throughout the COVID-19 pandemic. Companies will accelerate their adoption of AI technologies to optimize supply chains to gain competitive advantages on cost and customer service.

Many of the great successes of AI in 2021 will not be obvious because many developments will not be consumer facing

The AI-driven improvements will be embedded in our daily lives and operate behind the scenes with a renewed focus on areas such as production and supply chain optimization.

For businesses, things will run more efficiently, lower costs, reduced size of inventory and be more profitable. As a customer, we’ll notice that we receive an item faster or get it slightly cheaper.

Some key areas of AI development to watch in 2021 will be in Deep Reinforcement Learning (DRL), limited intelligence robotics, and computer vision.

The “data fusion” trend will accelerate in 2021, particularly in healthcare, government, and manufacturing

Cloud technology and adoption in 2020 set the foundation for businesses to do more with data. However, combining disparate data sources is still a pre-requisite for many analytics use cases. That’s where data fusion comes in. Data fusion helps create a more complete picture of a customer, business, or situation to answer questions, find new opportunities, and improve operations.

Last year, we saw this accelerate in marketing analytics for retail. Take Campari Group for example. This customer leveraged Microsoft Dynamics 365 platform and CustomerIQ to break down data silos and gain new customer insights during the pandemic.

We predict a similar acceleration in healthcare, government, and manufacturing this year. The healthcare industry will be able to take advantage of the FHIR standard for data interoperability (and PHI protection). Government agencies may find use for data fusion in areas like epidemiology and public health, where modeling can be limited by siloed health data. There are opportunities in manufacturing as well, such as collecting telemetry and improving reporting for faster recalls, maintenance programs, and to understand how issues in manufacturing can lead to issues in the field.

Businesses that succeed in transformation with AI will be the ones focusing on robust improvements to operationalization

The pandemic accelerated cloud migration and adoption in 2020, paving the way for new analytics and AI capabilities. However, the importance of operationalization remains.  The “build it and they will come” approach will not work with bigger data sets and more complex pipelines.

Successful businesses will be the ones who focus on building a solid strategy and roadmap to tackle critical use cases, who cultivate a strong data culture to improve end-user integration. DevOps and MLOps principles will also play a large role in improving operationalization in 2021.

Customer buying behaviors will continue to prioritize online and delivery in retail, shifting focus to fulfillment and customer experience

COVID-19 restrictions drastically changed customer buying behaviors, and the retail industry will continue to see customers favoring online shopping and delivery options. Retail is not dead, but it is changing and the focus will shift to improving customer experience and fulfillment.

Restaurants will need to effectively operate quick serve locations in a delivery fashion, fulfilling orders in a way that provides a good experience while remaining cost-effective. Intelligent Order Sequencing solutions that optimize customer order prioritization with machine learning will play a large role in this shift as businesses find themselves juggling mobile, online, in-store, and drive-through orders.

“Intelligence in the cloud” will morph to “intelligence everywhere” as AI workloads shift closer to the point of data collection to enable new scenarios

Edge hardware developments and the use of GPUs next to sensors (specifically cameras and microphones) for visual inference and voice transcription at the edge will open up new architectural patterns and scenarios that take advantage of the new capabilities.

We now have the infrastructure for visual inference and voice transcription at the edge, making it far more cost effective and commercially viable for use retail and quick-serve restaurant spaces.

Developments in edge hardware will also “open up” remote locations, such as mining and construction sites in the energy industry, where large amounts of data can’t be transmitted wirelessly. Visual inference at the edge could then be used for remote monitoring in conjunction with text alerts, to improve security and safety.

AI and ML training workloads will become a deliberate focus area for organizations already leveraging edge intelligence

Recent advancements in edge hardware have provided powerful edge gateways, creating opportunities for new usage patterns. More edge devices are being outfitted with GPUs, enabling GPU-based training for things like inference at the edge.

Inference took off in 2020, and we expect to see more solutions building off of that momentum. There will also be the cost benefit here. The expense of sending high-frequency data, such as video for visual inference, to the cloud for training will make pushing those AI and ML workloads to the edge an easy decision point.

This is likely to be seen in remote connectivity scenarios where there are increased limitations on bandwidth and throughput, as well as the need for near real-time insights.

Private 5G growth will primarily be driven via mobile edge computing initiatives for Industrial IoT scenarios

IoT is already exploding with a push to have everything connected and represented as a digital twin. 5G will be yet another mechanism for connecting to data sources and applying edge intelligence.

Private 5G creates the opportunity to ingest and process higher data volumes from a wider variety of devices, creating a great opportunity for industrial scenarios in manufacturing, drilling, or other sensor-driven use cases.