Listen to the audio version of the article
Artificial intelligence already exists inside homes. Here I am. The latest edition of the Internet of Things Observatory of the Polytechnic University of Milan tells us that more and more Italian families are adopting smart technologies to control the lighting, heating and security of their homes, for reasons mainly linked to comfort and the possibility of save on your energy bill. The question therefore arises immediately: how much will next-generation AI help consumers further raise the level of experience at home and optimize the use of connected devices?
The topic is open and the scenario described to Sole24 Ore by Gianpiero Morbello, Head of Brand & IOT at Haier Europe, is quite clear. “Artificial intelligence is a mantra, it is the hidden engine that makes the house work and is already integrated into the various household appliances to control their functioning and increase their efficiency, using dedicated development libraries and no longer just classic algorithms. Gen AI is an important topic and will progressively enter the smart home to continuously improve the experience of using the smart product, for example by generating new personalized recipes to cook.” Already today, ovens (such as Haier’s Bionicook ID Series prototype) are able to recognize foods and mix food information with the user’s cooking habits to prepare the dish in complete autonomy and generative intelligence should be seen as an additional piece of the personalization that the connected home is aiming for. “Consumer interaction with the smart devices of the future – Morbello warns, however – arises from adequate preparation of the data sources to be processed, from people’s skills and from the careful assessment of the risks that the massive adoption of artificial intelligence entails”.
The white goods industry is therefore experiencing a phase of great evolution, yesterday marked by the Internet of Things and today driven by AI (it is no coincidence that AIoT was born, and by machine learning models that must interpret the enormous quantity of data collected by the sensors installed in the various devices connected to the Internet. Intelligent appliances, this is the prediction of the experts, will largely replace the market of traditional household appliances with respect to an approach that requires innovation to be simple, accessible and versatile, as well as safe and sustainable. For Andrea Crociani, Marketing Manager AIoT Products of Xiaomi in Italy, the true added value of connected devices “lies in the interoperability capabilities of the device within an integrated ecosystem to which today a further piece has been added: l “artificial intelligence. The Artificial Intelligence of Things platforms represent the foundations of the house of the future and will be further remodeled with the contribution of generative AI to allow technology to be increasingly focused on dialogue with people”. The reference goes, of course, to virtual assistants. The most advanced models of understanding and generating natural language, in the vision of the Chinese house, will allow us to ask increasingly structured and complex questions, leading to actual conversations (this is the objective of the XiaoAI assistant, currently in the development phase testing in China) and to guide users in using the devices.
If the quality of human-machine interaction is destined to change, will the smart home with ChatGPT be more intelligent? Asking this question to the OpenAI bot, the answer was obviously affirmative and sufficiently motivated: it improves the quality of communication and the precision of the information provided by the virtual assistant, increases the analysis capacity of the data collected by the sensors, optimizes device management connected, helps the smart home become increasingly efficient and personalized as you interact with it. Generative AI, taking concrete examples, can understand more complex and flexible commands – “Tell me a way to reduce energy consumption” – rather than simply turning an appliance on or off. From a technical point of view, moreover, any implementation of Large Language Models can find a place in a control interface or a voice assistant inserted in a Matter-based ecosystem. A universal protocol-compatible smart home device, in other words, could then include GPT-3 or GPT-4-based functionality to answer questions and control other devices within the network. And identical features can be used to develop even more intelligent voice assistants, marking the development of the services of giants like Amazon. Michele Ravetta, Product Manager of Alexa, observed in this regard how today we are “at the beginning of a technological revolution that will continue for decades. We have been working on our LLMs for several years and today we are building new, much larger, generalizable models that will lead to accelerating the path towards an even more proactive and conversational interaction with the virtual assistant.” Alexa, coincidentally, already benefits from a large language model (called Teacher Model) which has improved the machine learning systems underlying the solution. But we are only at the beginning.