GuShiio.com Wen| Wang Zhiyuan
Recently, when chatting with friends, I can always hear the word spatial intelligence.
But when asked what it meant, everyone began to be sloppy. People who understand better get confused when they explain it, and they get stuck in it; people who don’t understand will feel dizzy when they listen to professional terms.
Therefore, I want to share “spatial intelligence” in the simplest way.
01
Spatial intelligence, to put it bluntly, is the ability to play with space in our brains. Imagine an object and you want to see what it looks like. Do you have to turn it 360 degrees to know?
Also, when driving to look at the map, should we first navigate to the destination, then zoom out on the map, and see the entire context before starting? This is especially true for renovating a house. You have to look at the top view from God’s perspective to find out if there is any problem with the layout. These are all done through spatial intelligence.
Simply put, spatial intelligence is our ability to handle spatial relationships. If I wanted to explain it clearly to others in a simple way, I would say: Spatial intelligence is like the process of restoring an object in 3D in your mind.
Recently, Hangzhou’s Six Little Dragons have become popular, and the concept of spatial intelligence has also become popular. So where did it come from?
It is also interesting to talk about its history. In ancient times, ancestors did not have navigation tools. They relied on their brains to remember pictures and their eyes to know directions. They had to catch up with prey when hunting and found their way home. These were all inseparable from spatial thinking.
Later, humans began to build houses and draw maps, and spatial thinking became increasingly important.
In 1940, psychologist Edward Chace Tolman proposed a theory: the cognitive map. It mainly studies how people and animals find their place in space.
By 1983, American psychologist Howard Gardner listed spatial intelligence as one of the nine major human intelligences in his book “The Framework of the Mind”. He said that spatial intelligence is not just about identifying the way, but also related to the ability to draw, design, and solve problems.
So it wasn’t until the 20th century that scientists began to study it, and since then, space intelligence has a scientific definition. However, the development of spatial intelligence does not rely on building a car behind closed doors. Technological progress and social needs are the “boosters” of its growth.
Let’s talk about technology first, which has probably gone through three key stages:
- Ancient times. With maps and compasses, humans can find out the direction on earth at once, and their navigation capabilities can leap directly.
- During the Renaissance. The painter used perspective methods and the architect began to design various complex buildings. At this time, people’s imagination rose to a higher level.
- modern. After the advent of computers, the entire situation was rewritten. From design software (such as CAD), geographical information systems (GIS), to virtual reality (VR) and autonomous driving technology, spatial intelligence became the core support behind it.
Every advancement in technology takes spatial intelligence a big step forward.
Let’s talk about social needs. In order to survive, ancient people relied on maps in their minds to hunt and find their way in complex environments. This was their instinct. In the industrial age, people had to rely on precise spatial planning and design to make a living in order to build houses and build roads.
Now, the urbanization process is accelerating and spatial planning is becoming more and more complex. Careers such as architects, engineers, and urban designers are all inseparable from spatial intelligence.
To design a high-rise building, the team must first build it on the drawings to see if the overall framework is reasonable; to plan a subway line, we must consider how to pass it underground without conflict with other facilities.
Not only these professional fields, but daily life is also inseparable from spatial intelligence. Using navigation to find the way and view decoration renderings, these seemingly ordinary activities are actually testing your spatial capabilities.
More importantly, spatial intelligence is particularly closely related to the current popular STEM fields (science, technology, engineering, mathematics). Research has shown that people with strong spatial abilities are more likely to achieve results in mathematics and science.
Therefore, the education community is also paying more and more attention to cultivating children’s spatial thinking abilities from an early age. By understanding this, you will understand why spatial intelligence cannot be clearly explained in one sentence.
Because if we only say that it is used to look at design, it will inevitably be a bit sloppy; if we say that it is not a single skill that runs through life, technology, and careers, it will inevitably be too conceptual. Therefore, we must start from the root causes, technological development and application in life to understand it.
02
Understand the ins and outs, dig deeper and see how it works.
The core logic of spatial intelligence is very simple to understand: the system first understands the space, and then completes the task. Specifically, a system must first know what it is like around it, then figure out how to handle it, and finally perform the task. There are four steps behind this:
Spatial perception: The system uses sensors to see the surrounding environment, such as self-driving cars detecting pedestrians and roadside obstacles in front of them.
Spatial representation: Convert perceived information into maps that the system can understand, such as generating a 3D model of road conditions.
Spatial reasoning: The system conducts analysis and decisions based on maps, such as planning a path to avoid obstacles.
Execution: Convert decisions into actual actions, such as driving a car along a planned route and moving a robot according to plan.
These four steps are intertwined and are indispensable. Key technologies to achieve these four steps include: sensors, computer vision, AI algorithms, and 3D modeling and geographical location information systems (GIS).
Wait. These concepts and technologies are not simply stacked together, but are interconnected like an assembly line.
Imagine you have a self-driving car. It has to use sensors to know if there are people in front and trees on the roadside, then turn the road into a 3D map, then figure out a road that doesn’t hit the tree, and finally press this road. Drive. The same goes for the sweeping robot. It first senses the surroundings, then plans how to get around the sofa, and finally starts cleaning.
Only when everyone performs their duties and cooperates closely can spatial intelligence really move.
By the way, there are two interesting things here: First, blind people also have the ability to spatial intelligence; they can build accurate spatial cognition through touch and hearing; they can judge the size of a room by the echo of footsteps, and even feel the layout around them.
The other is, don’t think that spatial intelligence is only useful in high technology. For example, Gaode maps, cameras, street views photographed, road conditions calculated by GIS, and route suggestions given by AI are all doing the work of spatial intelligence. So, this technology is not far away from us at all.
So, what are the differences or unique advantages compared to artificial intelligence?
I think the biggest difference between them lies in what they do and how they do it. The core of spatial intelligence is to deal with spatial problems, while traditional artificial intelligence is more like an honest person who is step-by-step, good at completing tasks with clear rules, such as playing chess and accounting.
They also have the same thing: they all have to rely on data for a living. However, the data sources are different.
Traditional artificial intelligence mainly relies on human-annotated data, such as tagged pictures in image recognition, or manually entered training data. Spatial intelligence relies on sensors to obtain information directly from the real world, such as data collected by cameras, lidar and other devices.
Then look at perception and autonomy.
Space intelligence relies on cameras and lidar to see the surrounding environment in real time and draw maps by itself. Like a robot entering a new room, it can figure out by itself where the walls are and where the doors are. Traditional artificial intelligence does not have this ability. It has to give it maps or manually marked information in advance before it knows what to do.
Also, spatial intelligence can figure out its own patterns from a large amount of data.
It can automatically optimize navigation routes based on traffic flow. Traditional artificial intelligence is not good. It relies on people to write the rules first, which is much less flexible.
More importantly, spatial intelligence can not only think, but also act. It can direct robots to grab objects or control character movements in VR games. These tasks may be difficult for traditional artificial intelligence.
Moreover, spatial intelligence has a strong ability to process complex data. Just like a navigation App, it can automatically adjust routes based on real-time road conditions. If traditional artificial intelligence wants to do this, it has to let people break down tasks and design them step by step. The efficiency and autonomy are much worse.
Therefore, if we use a metaphor: spatial intelligence is like a flexible space explorer, while traditional artificial intelligence is more like a plan executor who follows the rules.
03
Find out what is powerful about spatial intelligence, and then see in which fields it is used in and which problems cannot be solved by traditional methods.
Autonomous driving is the most familiar application scenario to everyone. Today, all major car companies are competing fiercely in this field. In fact, the application of spatial intelligence goes far beyond that.
Oculus’s VR equipment uses spatial intelligence to allow the screen to follow the user’s movements and line of sight changes in real time. When you turn around, the picture is immediately adjusted synchronously, and the immersion feeling is instantly full. However, traditional rendering technology not only has slow calculation speed, but also has stiff effects. Spatial intelligence makes this process smooth and natural.
There is also the construction and real estate industries.
In the past, architectural design and construction relied mainly on manual measurements and drawings, which was inefficient and prone to errors. Now, spatial intelligence can not only optimize design plans through three-dimensional modeling and real-time monitoring, but also make the construction progress clear at a glance, greatly improving efficiency.
Group nuclear technology, which has become very popular recently, is a typical representative of the application of spatial intelligence to the fields of design and construction. But it cannot be defined simply in this field. Because it has data in a lot of space, it can do more.
For example: helping e-commerce companies do virtual photography.
In the past, traditional e-commerce photography required building real-life sheds and employing professional photographers and models, which had a long cycle and high cost. Now, with the help of spatial intelligence software platforms, merchants can directly generate realistic 3D effects, switch scenes and backgrounds with one click, and generate high-quality product maps and videos in a few minutes, greatly reducing costs.
This is not an illusion. Amazon and FedEx are both using spatial intelligence to optimize logistics efficiency, and hospitals are also using AI to help people watch movies.
Some time ago, I went to experience skin care. Before nursing, the staff used the device to monitor my skin, and then I saw detailed information such as the condition of pores directly on the software. This is essentially a way for space intelligence to collect data and provide services.
Therefore, spatial intelligence is like a bridge that completely moves the real world to the virtual world. It has huge potential and may even completely change the way we interact with the world.
04
Of course, although spatial intelligence is powerful, it also faces some challenges at present.
The first is data collection and utilization.
In the past, when decorating houses, designers would use tools to measure the size and layout of the room and then record them in the software. Will there be a possibility in the future for AI to directly help us record this data?
I have a Tmall elf at home. If I could just say to it directly: Where did I put this thing? rdquo; Let it help me write it down, and ask it next time I forget. That would be convenient. But this also raises a key issue: privacy.
If I hand this data to AI, will it be leaked?
The same problem arises in other areas. Nowadays, many data inputs still rely on manual operations. Is it possible to automatically collect data through human-computer interaction in the future? This is not only related to technical realization, but also involves ethical and legal considerations.
When it comes to interaction, it is inseparable from AI hardware.
Like cameras, lidar, and even microphones, these sensors have indeed improved their functions when added with AI, but they have also brought new problems: high cost, complex maintenance, and easy performance to be affected by failures.
For example, if the camera of a sweeping robot is dirty, it may not be able to see obstacles clearly, resulting in a great reduction in work efficiency. This kind of problem makes the promotion and practical application of products more difficult.
To be honest, if I were asked to buy an AI microphone now, I might hesitate. After all, it took only a few days for a new generation of products to come out again. Who would want to spend money?
For complex tasks, computing power is a bigger bottleneck.
Most spatial intelligence products require strong computing power to process massive amounts of data. I noticed that there is a trend in China: GPUs originally deployed locally are migrating to the cloud.
This approach improves performance, but it may also increase latency and make users rely more on network stability; how to find a balance between performance and practicality remains an open question.
The most critical point is that spatial intelligence is currently fighting independently in many fields.
Because it has been implemented in the industry, it involves the cross-use of multiple disciplines such as computer science, mechanical engineering, and psychology. Therefore, it is not easy to achieve deep integration of fields. After all, talents from different professional backgrounds need to work closely together to break down disciplinary barriers.
Despite this, the future of spatial intelligence is still worth looking forward to.
The development of technology has never been achieved overnight. As long as breakthroughs continue, it will surely bring more surprises in the future. I hope these contents can help you have a basic understanding of spatial intelligence.
