In September 2024 I attended at the The Things Conference in Amsterdam. There, I saw many interesting talks about IoT, LoRaWAN, and related things. One of the talks that caught my attention was one where Jenny Plunket (Edge Impulse), and Matt Hatton (Transforma Insights) discussed the use of generative AI in IoT.
Let’s explore the interesting insights revealed during the talk.
What Do Generative AI and IoT Bring to the Table?
Generative AI creates data, images, or audio and trains machines to think creatively. IoT, on the other hand, connects devices to share and analyze data from the physical world. Together, they offer endless possibilities. From detecting anomalies in factories to enabling smarter robotics, they’re improving systems across industries.
As Edge Impulse’s Jenny Plunkett shared, AI models for IoT allow devices to “think” and respond intelligently—from microcontrollers to GPUs. Whether for detecting faulty products or spotting hazards, these models are making IoT deployments more effective.
Real-World Use Cases That Are Changing Norms
The panel at the conference shared fascinating real-world examples of how generative AI complements IoT deployments:
Smarter Robots in Factories
Ever heard of a robot dog sniffing out factory anomalies? Edge Impulse is working on this concept with NVIDIA’s AI platform, Isaac Sim. Smaller, cheaper robots equipped with sensors and cameras can perform tasks like counting pallets, detecting debris, or identifying problems. By combining on-device AI with cloud-based intelligence, companies can streamline operations while lowering data costs.
Synthetic Data for Training Models
Training AI models for IoT often requires loads of data. However, real-world data can be hard to gather or impractical. That’s where synthetic data steps in. Whether it’s generating audio of glass breaking to train security systems or creating virtual environments to test robotics, synthetic data accelerates model development. It’s an excellent first step before fine-tuning with real-world data.
Video Analytics for Smarter Monitoring
The rise of AI video analytics is transforming how businesses interact with physical spaces. For example, enterprises use AI-powered cameras to monitor customer behavior in retail environments or track worker safety in factories. These systems reduce reliance on traditional IoT sensors while providing deeper insights.
Challenges and Opportunities in Merging AI and IoT
While the potential is vast, there are hurdles to overcome. One issue is balancing on-device processing (for speed and privacy) with cloud computing (for scalability). The choice depends on the use case. For factory-based robots, it might mean sending only critical data to the cloud. For personal security, keeping everything local ensures privacy.
With the growing presence of large language models like ChatGPT, the industry is exploring its role in IoT. However, these models are massive and designed for general-purpose tasks. Training smaller, domain-specific models for edge devices could solve this scalability issue in the future.
What About AI Regulation?
Governments are stepping up regulation around AI. The EU AI Act enforces rules about documentation, oversight, and ethical use. While regulation hasn’t restricted many IoT applications yet, companies are advised to plan ahead. Working with legal advisors ensures compliance without slowing innovation.
Beyond laws, conferences like “AI for Good” are fostering dialogue about the responsible use of AI. They highlight the importance of designing systems that prioritize safety, fairness, and privacy.
Should Companies Build or Buy AI Solutions?
One common debate is whether to develop AI models in-house or rely on third-party tools. According to the panel, buying off-the-shelf models often saves time and money, especially for general-purpose AI needs. However, creating custom, lightweight models for specific IoT environments—like robot dogs or industrial monitoring—can deliver better results.
The Future of Generative AI and IoT
The combination of generative AI and IoT isn’t a futuristic idea—it’s happening now. We’re at an exciting crossroads where edge devices are becoming smarter, cloud resources are more accessible, and synthetic data pushes boundaries.
The next leap? Smaller, edge-optimized models that minimize cloud dependency but still offer robust intelligence. With continuing advancements and smarter regulation, the possibilities for generative AI and IoT seem limitless.
Conclusion
Generative AI and IoT are proving to be a powerful combination, transforming industries and solving problems in innovative ways. From smarter robots in factories to synthetic data driving AI training, the partnership between these technologies is creating real value. If you’re considering leveraging IoT and generative AI in your operations, now’s the time to dive in. Careful planning, thoughtful use of resources, and keeping an eye on regulations will help ensure success.
Ready to learn more? Watch the full discussion from The Things Conference 2024 above.
0 Comments