Follow and star “Everyone is a Product Manager“
Delivered daily at 07:45 AM
One reason for writing this article is Jason.AI’s “Product Perspective: Trends and Endgame of AIOT”. This article precisely analyzes the trends and stages of AIOT in terms of collaboration and interaction, which is very impressive!
However, many people mistakenly interpret certain “smart” communities’ “access control, video intercom, elevator control systems, and parking management systems” as AI homes, which is clearly a misguidance—true smart homes are much more complex and technological than this.
Author:wweiru
Cover image from licensed stock photo site Tuchong Creative
The full text is 5617 words and 3 images, requiring 11 minutes to read.
In the last semester of my university career, I interned at a company that specializes in smart homes (before the internship, my chosen thesis topic was the design of a smart home system based on PLC, which later changed to a smart home system design based on Zigbee), and after graduation, I stayed there, having spent nearly 3 years from internship to departure.
During this period, I worked on multiple Zigbee smart home projects (ranging from small apartments to large factories), visited several well-known smart home companies in China (such as Orvibo, Unisoc, Nanjing IoT, and Kedi), and interfaced with smart curtains, smart lighting, smart security (including surveillance), smart audio-visual (including background music), and smart locks.
In the last year at the company, I participated in a second startup (developing a Zigbee-based smart home system), which ultimately failed, leading me to leave the company and completely transition to a product manager role. Therefore, my deepest understanding of the IoT industry is in smart home systems.
1. Stages of Smart Homes
Smart homes are a very popular sector in the IoT industry. Since Apple released Homekit and Google acquired Nest, the smart home market has exploded.
The smart home industry can be divided into three stages based on device connectivity: single device connection, inter-device interaction, and platform integration; currently, the smart home industry is in the second stage.
First Stage: Single Device Connection
This stage saw a surge of single devices, which are more willingly referred to as smart hardware rather than single products.
In the era of smart single devices, Xiaomi is a leader.
For example, Xiaomi has successively launched the Xiaoyi camera, Xiaomi door/window sensor, Xiaomi alarm, Xiaomi speaker, and Xiaomi bulb; other manufacturers have introduced air quality monitors (for weather forecasting). The stickiness of smart homes in this stage is very low, and the initial novelty fades after a few uses.
Common single devices include: smart lights, smart locks, smart speakers, smart plugs, smart refrigerators, smart curtains, smart washing machines, smart air conditioners, smart rice cookers, and smart vacuum robots.
This stage is purely about connecting people to devices.
Second Stage: Inter-Device Interaction
In this stage, companies integrate all their single devices to enable interaction between products. For example, when a smart lock is opened, the lights automatically turn on.
In addition to self-integration by companies, smart home integrators can utilize an open platform from a company to integrate other third-party products into that company’s platform, providing customized interaction scenarios for end users; this stage has achieved partial inter-device connectivity through the efforts of certain manufacturers and integrators.
For instance, all single devices under Manufacturer A can be integrated into a single app, or a certain integrator can combine products from multiple companies into one system. The former is primarily represented by Xiaomi, whose app can control most of its single devices. The latter is represented by integrators like Orvibo and Control4, which integrate their products with those from other companies into the systems they develop.
Third Stage: Platform Integration
This stage aims for compatibility among products from different companies based on unified standards, but it has not yet developed to this stage. For example, the gateway from Company A can control the lights from Company B, and the sensors from Company C can command the vacuum robot from Company D to clean.
This stage requires true interconnectivity, not relying on a specific integrator or manufacturer, but achieving it through a certain protocol.
Currently, there is no universal protocol or platform that can achieve interconnectivity among smart home products. While Wi-Fi and Bluetooth are globally used, they still cannot dominate the market due to their inherent limitations (the former has high power consumption and limited device support; the latter has only recently supported mesh networks and is not yet widespread).
It is important to note that single devices and inter-device interaction do not strictly progress in chronological order. When I first entered the industry, there were already smart home systems with inter-device interaction, and it was later that Xiaomi brought smart hardware to a new level of popularity.
2. Current State of Smart Homes in China
Currently, the smart home market in China is polarized: on the manufacturer side, it is bustling with activity, as real estate companies, home appliance companies, internet companies, and AI companies are all entering this industry; however, on the consumer side, it is much quieter.
Relevant data shows that the penetration rate of smart homes in Europe and the United States has exceeded 35%, in Japan and South Korea it exceeds 25%, while in China, this figure has not yet reached 5%.
1. Consumer Challenges
Decision Difficulty
It is well known that the purchasing decision process for hardware products is longer and more complex than that for internet products, and once a certain hardware is chosen, the replacement cost is very high. This leads to a very high decision complexity for customers when choosing smart home products.
These issues cannot be changed through market guidance and education; even though many companies provide smart home showrooms, it is still very difficult to improve the conversion rate.
High Prices
Another point: the prices of smart home products are still very high.
Although domestic products are not expensive, the overall price is not low, and imported products have higher unit prices, leading to a decrease in customer purchasing desire, turning smart homes into toys for a select few.
Limited Consumer Exposure
Some smart home products in China are sold through internet channels, but ordinary consumers have few opportunities to be exposed to them.
Moreover, the product information obtained through online reviews often differs significantly from the actual product experience, making many consumers feel as if they are looking at the moon from afar, seeing it but unable to touch it, resulting in relatively low purchase intentions.
Not Smart Enough
Another reason consumers are not buying is that the intelligence in consumers’ minds differs greatly from what the market offers.
For example, many communities are equipped with access control, video intercom, elevator control systems, and parking management systems, and then claim to be smart communities.
—These are far from what consumers envision as smart homes, leading them to preemptively think that smart homes are just a facade.
2. Manufacturer Challenges
There are many manufacturers in the smart home sector, with domestic giants like Alibaba and Huawei (including Honor) also having smart home products.
Overall, manufacturers can be divided into those represented by Xiaomi, which originated from smart hardware/single products (first making single products, then creating a super app to manage all single products, and finally evolving into a system), and those represented by Orvibo, which are smart home system manufacturers (building their own systems/platforms and then integrating third-party products).
3. Future of Smart Homes
1. Multi-Modal Interaction
It must be acknowledged that smart homes differ from other AIOT industries in that they require multi-modal interaction methods and multi-modal collaboration (between devices)..
The following paragraph is from Qiling:
Multi-modal interaction refers to the combination of various interaction means, such as integrating multiple senses, including text, voice, vision, motion, and environment. Humans are a typical example of multi-modal interaction; in the process of communication between people, expressions, gestures, hugs, touches, and even smells all play irreplaceable roles in information exchange. Clearly, human-computer interaction in smart homes must involve more than just voice; it requires parallel multi-modal interactions. For example, if a smart speaker detects that no one is home, it does not need to respond to a wake word mistakenly played on the TV and can even switch itself to sleep mode; if a robot senses that its owner is watching it, it may proactively greet the owner and ask if assistance is needed. Multi-modal processing undoubtedly requires the introduction of joint analysis and computation of various sensor data, which includes one-dimensional voice data as well as two-dimensional data such as camera images and thermal images. The processing of this data necessitates local AI capabilities, thus creating a strong demand for edge computing.
2. Identity Verification
Currently, smart home systems generally cannot recognize you as you, while some smart single devices can.
Currently, identity can be confirmed through smart locks or cameras, but smart locks are only at entry points, and cameras are generally used for perimeters and entry points (although some clients have previously requested the installation of numerous surveillance cameras indoors). However, how can identity be verified indoors?
Humans typically recognize others through auditory and visual cues.
In smart home systems, there are several devices capable of recognizing you, such as TVs with cameras and smart speakers.
There are many scenarios related to identity verification. For example, when you return home, background music or smart speakers, including lighting, will adjust according to your personal preferences, based on your daily behavior, gender, preferences, return time, weather, and specific dates (birthdays, anniversaries, etc.) through self-learning, aiming for personalization.
Another scenario related to security.
When the security system is armed at night, if you get up and move around, the smart home system can recognize your activity, and the security sensors in your vicinity will automatically be disabled. When you leave that area, the smart home system will re-enable them—this avoids the need for you to manually arm or disarm the system.
3. Location Awareness
The aforementioned security scenario is actually related to location awareness, as smart home systems not only need to recognize you but also need to know where you are (bedroom, kitchen, living room)..
One of my personal ideas is to endow smart home systems with 3D modeling capabilities, allowing them to understand the entire home structure, which can then be assigned to other devices, such as vacuum robots or smart speakers that may move.
There are many related scenarios; for example, when you are cooking in the kitchen while listening to music from a smart speaker, and the smart speaker suddenly interrupts with a voice saying, “There is a stranger at the front door,” while displaying the monitoring image at the door. You see from the monitoring image that it is your neighbor or a friend, or even a delivery person. You tell the smart speaker to open the door for them or say, “Delivery person, please leave the package at the door; I will pick it up shortly.” At this point, the smart speaker will relay this message through the doorbell (assuming it has voice functionality) to the person waiting at the door.
Similarly, if you are watching TV in the living room, a small window will appear on the living room TV displaying the door monitoring, and the smart speaker in the living room will announce, “Someone is at the door,” while other TVs and smart speakers in different rooms will not issue corresponding prompts; even when multiple TVs and smart speakers are in use, the smart home system can determine which TV or smart speaker should issue the prompt (for example, prioritizing young people, followed by the elderly and children)..
If no one is home, then when someone rings the doorbell, neither the TV nor the smart speaker will provide a prompt about someone visiting.
4. Connectivity
Smart homes connect people with things, while smart communities connect services with people and various smart home systems. Communities provide more personalized services to homeowners through various software and hardware facilities.
For example:
You have scheduled a meeting with a client at a certain location at 9 AM tomorrow. The smart system sets an alarm for you to wake up at 7 AM, arranges for breakfast delivery at 7:30 AM, schedules your departure at 8 AM (with a vehicle waiting at the community’s east gate), and plans for you to get in the car at 8:10 AM. The smart system selects a transportation method based on the distance from your residence to the destination, estimating a 40-minute journey. Since your vehicle is restricted, the smart system chooses Didi ridesharing based on your calling habits and selects a premium car (not a regular ride or taxi). Given that the community offers free pick-up services, the smart system informs the community to wait for you downstairs at 8 AM. This continues with the alarm settings—originally set for 7:30 AM, the smart system determines that you need to wake up at 7 AM to make the schedule, so it will disable the alarm set for 7:30 AM tomorrow (without affecting the one for the day after).
5. Centralization and Decentralization
According to the Nash equilibrium principle, an organization reaches a stable state where the decisions made by the group are optimal, and any other choice would disrupt this equilibrium. Human society actually has multiple Nash equilibria—sometimes leaning towards centralization and sometimes towards decentralization.
The Nash equilibrium tells us that the organizational form of a group will find a balance between a single center and decentralization, and there can be multiple such balance points.
So, in the smart home industry, is it centralized or decentralized?
The current trend is still centralized; even if in the future every device becomes smart and has edge computing capabilities, there will likely still be a central entity responsible for coordination.
Why do I say this?
If every sensor and terminal device is treated as an individual, and the home is viewed as a conference room where everyone is speaking at once, will there be a conclusion?
No.
This mutual command transmission is akin to the various sounds in a bustling market; thus, there will largely still be a center, while allowing for the existence of other centers. For example, if a sensor detects that the floor is dirty, it can send a command to the vacuum robot to clean.
Another example to illustrate:
When the robot detects that the owner is not home, it automatically enters sleep mode. The question arises: how does the robot know the owner is not home?
Method One: The robot circles through all rooms and finds no one home, then returns to its previous position and enters sleep mode, informing other devices to do the same.
Method Two: The robot shouts, “Has anyone seen the owner?” (broadcasting), and all devices tell it they have not seen the owner, or a specific device informs it that the owner is out (for example, a camera or speaker), prompting the robot to enter sleep mode.
Method Three: The smart home system informs the robot that the owner is not home, prompting the robot to enter sleep mode.
Methods One and Two can be seen as decentralized; Method Three is centralized.
6. Data Analysis
Recently, I watched a popular drama called “All is Well,” where there is a scene where Su Mingcheng notices that Su Daqiang goes to the bathroom several times at night, inferring that Su Daqiang has health issues.
So, if a smart toilet detects that a person has been going to the bathroom several times every night, can it infer that this person has health issues?
The current answer is: it cannot make that judgment.
The reason is that current smart toilets cannot analyze the data they generate and cannot determine the identity of the user.
However, some medical products already have basic data analysis capabilities, providing suggestions or assistance based on the user’s physiological state.
Thus, it can be inferred that smart hardware or smart systems must actively analyze the data generated by the device or system and provide relevant suggestions to users.
7. Privacy and Data Protection
The previous example mentioned about the smart toilet detecting potential health issues raises the question: if the smart toilet transmits this information to the manufacturer’s cloud platform and subsequently informs related pharmacies or hospitals, leading to the individual receiving numerous advertisements or products related to how to treat certain issues, or even receiving unsolicited push notifications.
Should such situations be allowed to occur?
Moreover, since smart devices or systems hold a large amount of personal data, and both smart devices and systems need to regularly communicate with cloud platforms, if the communication content is intercepted by malicious individuals, it could be exploited by those with ulterior motives.
Therefore, every manufacturer should take concrete measures to protect data—whether during data transmission or in the cloud platform where data is stored or generated.
We know that to achieve intelligence, a large amount of data is required; however, due to privacy and data protection concerns, smart hardware or systems cannot directly transmit personal privacy and data to a cloud platform. So how can model training be conducted?
Regarding Privacy:
Personal data is private, but the data characteristics of a group of people are no longer private. For example, if a person likes to buy smart hardware and 80% of men in a certain area also like to purchase smart hardware, the former involves an individual, while the latter does not involve an individual (it can be public). Therefore, personal data can be anonymized before being uploaded to the cloud, but this process must obtain the consumer’s consent.
For Model Training Approaches:
Edge computing can be utilized to conduct preliminary training of data locally, then return the results to the cloud platform for further training, ultimately deploying the model to the respective terminal devices or systems.
References
Jason. Everyone is a Product Manager. “Product Perspective: Trends and Endgame of AIOT”
Qiling. Zhihu. “AIOT? Artificial Intelligence of Things”
———————— END ————————
———— / Recommended Reading / ————
>> Product Managers Under Smart Homes
>> How to Design User Experience for AI Products?
>> Essentials: AI + Product Manual with Millions of Daily Active Users .doc
>> AI is Not Magic: What AI Can and Cannot Do
>> Three Factors for IoT Manufacturers’ Success: Content / Software / Products
❖ System Enhancements: The Four Essential Capability Matrix for PMD
❖ Product Director Course in Beijing on June 15, waiting for you to join
❖ The first 30 registrants can participate in a social dinner for in-depth interaction with the instructor for free
▼ Click “Read the Original” to view course details