As someone keenly interested in both home automation and generative AI, I’ve been paying attention to how the two might be combined. Both technologies have their limits. But like a composite, could combining them make both better? I think so. I also think there is one very important thing AI could do for home automation.
What is that one thing? Improving voice control. For my money, voice control is second only to geolocation control in terms of what I think home automation manufacturers should be focusing on as the next big thing.
Voice Control at Present
As things currently stand, controlling home automation devices with one’s voice can be done via phone apps and smart speakers. A smart speaker setup is preferred because it allows voice control without having to pull a device out of your pocket. And with smart speakers strategically located throughout a home, voice control is frictionless – until it’s not.
Voice control is not as frictionless as it should be because home automation systems still struggle with natural language processing (NLP). I know firsthand. My system struggles to recognize what I am saying if I don’t use just the right words and pay attention to how well I enunciate.
Needless to say, current voice control capabilities are spotty at best. Even when combining something like a Vivint automation system with a Google Nest smart speaker, the results are often less than what consumers expect.
How AI Can Help
So, how can AI help in this regard? AI he’s starting to have a real impact on improving NLP. And as large language models (LLMs) continue to grow and improve in their accuracy, they have the potential to transform how smart speakers are used. Google is already heading down that path.
In a recent blog post, Google shared how they are starting to integrate their Gemini LLM into Google Nest products. The blog specifically mentioned incorporating Gemini with the company’s smart speakers.
Although the blog post is short on details, it doesn’t take much to figure out what Google is working on. They hope to have LLM and NLP systems that are eventually mature enough to allow people to interact with their smart speakers the same way they talk to friends and family members.
Right now, users must use very specific commands to get their smart speakers to respond accordingly. Use the wrong words and your smart speaker will tell you it doesn’t understand. I have that problem with my system. The ultimate goal is to completely eliminate that barrier. AI could make it possible.
How It Could Happen
NLP seeks to understand human speech by analyzing words in relation to their context. Every word in a sentence is compared to every other word in terms of position and definition. From those comparisons, an NLP system infers meaning and intent.
The big challenge is that people don’t use words the same way every time. Meanings are different. Context differs. Even sentence structure and syntax can vary considerably among users.
AI offers the advantage of being able to analyze endless amounts of data, compare it, and respond accordingly. The larger an LLM grows, the more accurate it becomes. So, applying AI to NLP should theoretically make the latter more accurate over time.
I’m looking forward to seeing how Google and other tech companies transform home automation with AI. I’m guessing it won’t be long before voice control over home automation systems experience a significant boost. And when that happens, home automation is going to be a lot more pleasant to use.
