• IoT For All
  • Posts
  • How IoT will feed AI’s endless appetite

How IoT will feed AI’s endless appetite

Plus, this week's top stories from IoT For All

Author

IoT For All
February 13, 2025

In partnership with

Hello readers!

Welcome to this week’s IoT For All newsletter where we’ll be talking about how IoT is an excellent AI data generator, the finer points of Total Productive Maintenance and Overall Equipment Effectiveness, and whether data historians are holding back Industry 4.0!

Keep your hands where I can see them

This week, Meta has been in the news for torrenting terabytes of books to feed its large language models. But the company formerly known as Facebook hasn’t only been taking. Late last year, it released the HOT3D dataset, some 13 hours of first-person footage featuring hands just doing stuff. Both headlines share a root cause: AI developers are running out of data to train their models.

As early as 2022, research firm Epoch AI presented a paper suggesting that “high-quality language data” for training Large Language Models could be exhausted as soon as 2026. Elon Musk says it already happened. Of course, different models require different types of data as input, sources of which will be exhausted and replenished at different rates. But there’s no doubt it’s all being devoured. Solutions vary from attempting to tap previously private caches of unadulterated data to attempting to algorithmically create “synthetic data” that’s (hopefully) “good enough.”

What’s this got to do with IoT? Meta’s HOT3D dataset connects the dots. Captured in part with the company’s Ray-Ban-branded wearable cameras, the footage shows the potential for IoT to supply that much-needed data in spades. In this case, the data was produced for its own sake, and openly shared because Meta has an interest in improved hand-detection technology for its Metaverse aspirations. But it’s only a small hop to see how IoT sensors, specifically those deployed in enterprise applications, could be an ever-flowing fountain of data in addition to what they were already intended to do.

Vibration sensors intended to pick up alarmingly out-of-spec movement can supply the wealth of data needed to spot issues before they’re noticeable to the human eye. Location sensors intended for equipment or shipment tracking provide the logistics information an AI model requires to streamline warehousing locations and workflows. Security cameras, originally installed to provide human-reviewed video feed, can generate the oceans of visual data required to train computer vision models to spot manufacturing defects, optimize workstation placement, or handle compliance monitoring for PPE requirements, restricted zones, and more.

Some types of data may very well remain quite scarce. At least, not until temperature sensors develop the ability to write prose. But as the initial sugar rush of rapid-fire AI advancement burns off and the natural reservoirs of data dry up, enterprises with IoT-powered data-production factories already up and running will be well-poised to make whatever leap comes next.

📖 Top Articles

Productivity Is for Humans: TPM and OEE in the IIoT Era

Wired Magazine Founder Kevin Kelly once said: “Productivity is the wrong thing to care about in the new economy. Productivity is for robots [and machines].” But in today’s modern factory, he’s only partly right. Yes, productivity matters for machines, but it’s also the primary concern for many people who manage, maintain, and optimize machine productivity.

Meeting the multi-faceted requirements of the Cyber Resilience Act (CRA) is a demanding task for manufacturers. There are overlapping challenges in maintaining secure updates, compiling and updating a comprehensive SBOM, identifying vulnerabilities, remediating vulnerabilities, providing secure product updates, and ensuring a secure-by-default configuration. These elements do not exist in isolation.

People have been discussing Industry 4.0 for over a decade, but now it’s finally getting serious. The recent AI hype could be credited with reigniting the conversation. Thanks to extensive publicity of OpenAI, it’s commonly known that the more data you feed AI, the smarter it becomes. And who are the gatekeepers of data in the industrial landscape? Data historians.

🔥 Rapid Fire

Build Smarter, Faster: AI Voice Agents for Every Industry

Save time building your AI calling assistant with Synthflow’s AI Voice Agent templates—pre-built, pre-tested, and ready for industries like real estate and healthcare. Get started fast with features like lead qualification and real-time booking. You can even create and sell your own templates to earn commissions!

🎙️ The IoT For All Podcast

This week, we spoke to Fabrizio Del Maffeo, co-founder and CEO of Axelera AI, to discuss edge AI. The conversation covers the importance and benefits of edge AI, such as reduced latency, real-time decision-making, and privacy, optimizing algorithms and hardware for edge devices, the potential of AI in various industries, cloud computing, retrofitting existing solutions, and the impact of generative AI.