Most People's Understanding of AI is Wrong
-
In the new era, can artificial intelligence help with marketing or brand services?
Before reading, consider:
In which aspects can AI not surpass humans?
In what key areas can AI play a role in marketing or brand services?In recent years, I have been researching computational advertising on the internet. There is a lack of systematic introduction to this field in the market, so I even wrote a book on computational advertising, which has fortunately been recognized by the internet industry.
Today, I want to discuss a case with you: In the new era, can artificial intelligence help with marketing or brand services?
It is widely believed in society that artificial intelligence is simply machines' deep learning of algorithms. But I can accurately tell you that this concept is incorrect.
Artificial intelligence was proposed by the pioneers in the field of computer science with the aim of using computers to replace humans in solving certain intelligent problems, such as replacing human hearing, speaking, reading, writing, and communication. Artificial intelligence can achieve these through machine learning methods, but it doesn’t have to.
In other words, machine learning is a method and tool of artificial intelligence, not artificial intelligence itself.
In the narrow sense, the field of artificial intelligence mainly includes "perceptual intelligence" and "cognitive intelligence": perceptual intelligence replaces human senses, while cognitive intelligence replaces human thinking and communication.
Boston Consulting Group mentioned in a book that artificial intelligence can provide personalized services, advertisements, and interactions for customers in corporate marketing services.
All existing advertising and marketing efforts do not interact with users; they are like handing out flyers and then it’s over. But data shows: Brands that can create personalized experiences can increase revenue by 6%~10%.
The term "personalization" is familiar to everyone. Today’s advertising and marketing are personalized, tailored to each individual. But using artificial intelligence technology can make advertising and marketing interactive on the basis of personalization.
I’ve noticed that many people, when talking about AI marketing or intelligent services, always say things like:
"How many active devices we have," "How much data we’ve collected through SDKs," "How many tags we’ve built," "Identifying target audiences," "Precision targeting," etc.
These statements cannot be verified or falsified. I believe they use machine learning technology, but it cannot be said that they use artificial intelligence itself. The topic we are discussing today is using artificial intelligence itself to promote marketing services.
This process is quite difficult because most people’s current understanding of artificial intelligence itself is still at a very elementary stage.
All advertising and marketing today have one characteristic: proactive outreach. But it is mainly one-way communication, sending out ads and then leaving it at that. Such advertising and marketing severely lack communication and interaction, which is incorrect.
In developing marketing, a very useful technique is: targeted interaction. Using data and machine learning in this way allows flexible responses to different situations, delivering ads tailored to each individual.
Now let’s look at customer service. Unlike marketing, customer service is when users come directly with problems, but no research is done on the users; instead, issues are addressed as they arise.
These two tasks are essentially the same thing, and insurance salespeople do them best.
Insurance salespeople do not distinguish between marketing and service (customer service) processes. First, they communicate with you, become friends with you, with the goal of obtaining your information—a process they are very patient with.
In marketing, we passively obtain user information, while insurance salespeople actively obtain it, much like customer service, targeting collected needs for sales opportunities.
For after-sales service, insurance salespeople also care deeply. They often send small gifts during holidays, ask about the family situation, collect feedback, continuously track and update user information, with the ultimate goal of secondary marketing.
This seamless integration of marketing and service has two points worth learning: one is the service mindset, and the other is the awareness of customer data.
In the future, we might imitate the approach of insurance salespeople to establish a virtual marketing process. Of course, this will take time.
Current AI research is divided into four categories: perceptual intelligence, computational intelligence, motor intelligence, and cognitive intelligence.
Perceptual intelligence is the problem people have long wanted to solve the most. Perception includes eyes, nose, ears, speech recognition, etc. These have evolved the longest, starting from when animals first developed these senses.
For computational intelligence problems, AI can find quick solutions with its high computing power. This is very challenging, with Go being the most typical example. But relatively speaking, AI has already solved the Go problem early. Today, machines can outperform humans in all game theory problems.
Motor intelligence has also evolved the longest, starting from primitive times.
Cognition is a uniquely human ability, the source of reasoning, thinking, and communication. Our communication, dialogue, and understanding are all cognitive intelligence.
AI can surpass humans in game theory problems, while humans will evolve in perception, motor skills, and cognition.
In the field of marketing services, we have also explored AI research extensively—
I’ve experienced the interactive experience of outbound call robots. Chatting with them feels very natural, and unless you’re a professional, you can’t tell it’s a robot. This is a big step forward in the field of interaction.
Outbound call robots perform quite well in marketing, equivalent to an insurance salesperson recommending suitable products to you. This concept is "precision marketing."
However, robot customer service is the opposite of outbound call robots. Customer service is about service, while outbound calls are about marketing, but the logic of both is similar: using technology to solve marketing service problems.
In summary, the direction of brand marketing and services in the AI era is—establishing perceptual and cognitive abilities, creating service agents for each customer’s entire lifecycle, recording and analyzing user data, and providing interactive continuous service.
This agent must have a brain (something interactive), usable products, a personalized model (similar to an insurance salesperson’s service mindset, tailored to the customer’s specific situation), and customer data.
For example, KFC’s Colonel Sanders can completely become an agent in the AI era, with all marketing and service channels interacting with customers through this image.
Today’s machines can already actively recognize identities, though there are legal boundaries. For example, Focus Media’s screens can actively interact with people under reasonable authorization.
Although this scenario hasn’t been realized yet, we can imagine it.
Soon, machines in the field of perceptual intelligence will become increasingly close to humans, making them usable in proactive collaboration scenarios. But don’t think AI has reached a turning point—its fundamental problems remain unsolved, and AI’s capabilities in cognitive intelligence are still far behind.
Cognitive intelligence problems are very challenging; all inspiration comes from cognitive intelligence.
Our conceptual reasoning isn’t learned from school education. Just like a college student and an illiterate person can communicate basic concepts without any barriers. Over 80% of human concepts and logic are acquired before the age of six in a very奇妙 way.
Currently, all AI is weak AI. For example: this common sense—"I drank a bottle of water, so the bottle became lighter"—machines can’t learn. Or some news that humans can immediately recognize as fake, but machines can’t.
Because achieving this requires very complex and奇妙 common sense, and machines can’t find the corpus to learn this common sense. So, in terms of understanding and reasoning about concepts, machines currently cannot catch up to humans.
This also means our imagination should revolve around human-machine interaction in marketing and services. If it’s in a broad field, it’s very difficult.
However, in some vertical fields, like medicine, we feel it’s very professional, but for machines, it’s simple. This is because professional knowledge is all in books; as long as there’s corpus, machines can learn.
Therefore, cognitive intelligence is the biggest challenge for AI today. This isn’t something that can be solved simply by functional improvements or resource investment—it’s a theoretical gap.
iFLYTEK started with voice synthesis. Two years ago, it achieved machine simulation of human voices with remarkable realism.
Simulating visual appearances is no more challenging than voice simulation. iFLYTEK has a virtual host, currently still in image form, but this indicates that producing a lifelike human figure in the future won't be difficult.
iFLYTEK also aims to combine the core technologies of artificial intelligence with marketing and services, such as simulating a human salesperson for continuous marketing and service.
Data collection in the marketing field, combined with personalized capabilities, forms the basic human-computer interaction abilities of iFLYTEK. The next steps involve mobilizing data, voice recognition, visual and virtual image output. The only missing piece is AI's understanding, reasoning, and communication of concepts.
Therefore, under the premise of broad scenarios, creating intelligent agents for marketing will take some time. However, this is feasible in certain specialized vertical scenarios.
I hope to present more interesting and refined cases in our next discussion. Thank you.