Soracom has integrated artificial intelligence capabilities into its Internet of Things (IoT) products, ushering in a new era of device connectivity and communication. What does this decision mean, and will it have a lasting impact on cellular IoT services?
Soracom Introduces AI Capabilities to Its IoT Ecosystem
Soracom, a Japan-based global connectivity provider, is bringing generative AI capabilities to its internet-connected ecosystem. IoT-focused telecommunications companies can use any of the company’s three tools to process data in real time or connect devices to the machine learning tools of major hyperscaler platforms.
Each of the three services offers unique capabilities, so telecommunications companies can choose one based on their needs: To process, analyze, and transfer IoT data, they can choose between Soracom Query, Soracom Harvest Data Intelligence, and Soracom Relay.
Many companies are incorporating machine learning and generative models into internet-connected devices, and some are developing standalone, out-of-the-box software solutions for communications. But Soracom’s integration of AI into its cellular IoT services is novel in that it focuses on back-end operations, rather than automating management or customer-facing functions.
What can Soracom’s three new AI services do?
Soracom Query allows you to mine IoT-generated data using Structured Query Language (SQL) queries from your Customer Lifecycle Intelligence (CLI) or Business Intelligence (BI) tools without deploying servers, meaning you can pull information from a database without using separate storage – Soracom calls this “Managed Data Warehouse functionality.”
Soracom Query has an automatic data loading feature (incrementally processing incoming information without any additional configuration), so you can easily run complex analytical SQL queries, even on large datasets. If you’re interested, the service is available in technical preview.
Soracom Harvest Data Intelligence is a generative model (specifically, the GPT-3.5-Turbo algorithm provided by Microsoft Azure OpenAI services) that analyzes time-series information, i.e. it processes and interprets a series of data points recorded at intervals to identify notable trends, patterns, outliers, and anomalies.
You can tell the AI how to process the data and generate insights like if there are deviations in the time series. You can also give it commands like telling it to explain trends. The service is available in public beta.
Soracom Relay allows cameras running Real-Time Streaming Protocol and Real-Time Transport Protocol to send video and audio files on demand to cloud storage systems such as Soracom’s Harvest Files or Amazon Web Service’s Kinesis Video Streams. The service is available in technical preview.
Implications for the future of cellular IoT services
These three AI-powered products have the potential to have a lasting impact on cellular IoT services due to their significant benefits: Soracom Query, for example, allows telcos to dynamically scale BI and CLI SQL query processing, reducing infrastructure costs.
In addition, the ability to efficiently process massive data sets helps decision makers extract key insights faster, allowing them to quickly adapt their data mining and utilization strategies to meet evolving device needs, use cases and market changes.
Enhanced security is another benefit of the integration: Soracom Harvest Data Intelligence can alert you to potential cyber threats by flagging anomalies. Given that the average cost of a single data breach is $4.24 million, this service can result in significant cost savings that can spur innovation and accelerate research and development.
While some decision makers may not prioritize cybersecurity, malicious actors pose a major threat. In one case study, a telecommunications company lost approximately 33% of its enterprise value and approximately 100,000 customers after experiencing a data breach. Cellular IoT services capture and transmit millions of data points every day, so security is crucial.
One of the benefits of Soracom Relay is its cost-effective scalability, dynamically adjusting resolution, bitrate and video encoding based on network conditions, demand and server volume to optimize bandwidth and storage resources and greatly improve the efficiency of data acquisition and transmission.
It could also help companies identify anomalies in real time, for example by drawing attention to lost data packets and enabling retransmissions. At the very least, the activity provides an observable log so there’s no need to worry about data loss. The processing can also provide insights that can help retune streaming parameters such as bit rate, resolution, keyframe intervals, etc.
Scalability is one of the main implications of these tools, as they rely on autonomous models. Considering that there will be 30.9 billion Internet-connected devices by 2025, the ability to maintain a dynamic state for cellular IoT services is crucial. After all, the size of the dataset is positively correlated with the number of devices, so it is important to consider future transmission demands and storage flexibility.
Possible challenges arising from Soracom’s AI integration
Potential challenges that could arise from Soracom’s AI integration revolve around the typical pitfalls of machine learning technology, the main one being inaccuracy. Unlike other similar automation tools such as robotic process automation or scripts, algorithms operate autonomously and can make biased decisions or hallucinate information.
Telecommunications companies hoping to use AI to derive valuable insights or automate transmission processes may face errors because AI is often prone to mistakes. Bad actors can even intentionally manipulate AI. One study found that a machine learning model misclassified images more than 69% of the time due to adversarial examples (inputs meant to trick the algorithm).
As cellular IoT services expand, optimizing the efficiency of acquisition, processing, analysis, and transmission will be critical. While the impact of Soracom Query, Soracom Harvest Data Intelligence, and Soracom Relay will be mostly positive, over-reliance could cause problems. Enterprise leaders would be wise to thoroughly test the limitations and capabilities of these technology previews before adopting any one of them.
The Future of AI Technology in Cellular IoT Services
As the IoT and AI fields expand, ready-to-use services like Soracom will become more prevalent. Currently, their impact is mostly positive: they have the potential to optimize many back-end processes, reduce infrastructure burden, and improve insight-based actions. However, these tools should be approached with caution, taking into account that they are still in technical preview.
