Left Hand Workplace, Right Hand AI
With Stories, Methods, and Tools
January 20, 2025
Article 19 of 2025, Original Article 785
Full text 2682 words, reading time about 8 minutes
Since OpenAI launched ChatGPT on November 30, 2022—a technology known as Generative AI (AIGC, Artificial Intelligence Generated Content), the application of AIGC has started to grow explosively, as it brings great convenience to our daily work, life, and learning.
01 AIGC Seems to Be Omnipotent
In the workplace, AIGC can become a powerful assistant for content creators. For example, a copywriter can input key information such as product features and target audience into the AIGC tool when writing product promotion copy, and AIGC can quickly generate multiple versions of draft copy. These drafts not only provide abundant creativity and ideas but also save a lot of time in brainstorming and drafting. Additionally, for designers, AIGC can quickly generate design sketches and concept images based on given themes and style requirements, providing inspiration and reference for designers to complete design projects faster.
AIGC also has many applications in life. For example, when planning a trip, you can consult AIGC for information about the destination. It can provide you with detailed travel guides, including recommended attractions, food introductions, transportation guides, etc. Moreover, it can even customize a personalized travel itinerary based on your interests and time schedule.
Source: Kimi
Furthermore, AIGC can also play a role in the smart home field. By integrating with smart voice assistants, AIGC can understand user needs and provide corresponding services, such as checking the weather, playing music, controlling home appliances, etc., making life more convenient and comfortable.
In education, AIGC shows enormous potential. When students are writing papers, AIGC can assist them in literature reviews, grammar checks, and content optimization. It can also generate personalized practice questions and learning materials based on students’ learning progress and knowledge levels, enhancing learning efficiency. For language learners, AIGC can provide real-time language translation and speaking practice, helping them better master language skills. Additionally, AIGC can provide virtual tutor services on online education platforms, answering questions and providing guidance, addressing the issue of insufficient teaching resources in traditional education.
02 AIGC Is Not Omnipotent
“If you only have AIGC as a hammer, every problem looks like a nail that AIGC can pull out.”
However, just like artificial intelligence (AI) is not omnipotent, AIGC is not omnipotent either.
As shown in the figure below, in fact, AIGC is only a small part of the broader field of artificial intelligence, and solving most business problems requires combining different artificial intelligence technologies. If this fact is ignored, you may overestimate the role of AIGC and apply it in inappropriate scenarios. The consequence of this is that you not only fail to achieve the expected results but also waste resources, such as time and money.
03 How to Evaluate When to Use AIGC, When to Use Other AI Technologies, and When to Use Combined Technologies?
First, clarify whether using a certain artificial intelligence technology creates value for the business and whether implementing this technology is feasible. This is important because some scenarios are not suitable for using artificial intelligence and are not worth further consideration.
Below are some references for scenarios suitable for AIGC:
Very Useful: Content Generation, Conversational User Interfaces, Knowledge Discovery
Somewhat Useful: Segmentation/Classification, Recommendation Systems, Perception, Intelligent Automation, Anomaly Detection/Monitoring
Almost Useless: Prediction/Forecasting, Planning, Decision Intelligence, Autonomous Systems
If you cannot accept the risks that AIGC may bring and these risks cannot be effectively mitigated, then AIGC is not suitable for your scenario. These risks include unreliable outputs (serious nonsense), data privacy, intellectual property, liability, cybersecurity, and regulatory compliance, whether these risks are individual risks or combined with other risks.
04 What Other AI Technologies Can Be Considered?
For those areas labeled as “AIGC is not very useful,” other artificial intelligence technologies can be considered. It is wise to try a simpler alternative AI technology before delving into generative artificial intelligence; they are usually lower risk, lower cost, and easier to understand.
Common, mature artificial intelligence technologies include non-generative machine learning, optimization, simulation, rules/heuristics, and knowledge graphs. Emerging technologies such as causal AI, neuro-symbolic AI, and first-principles AI are also worth attention.
Non-generative Machine Learning: In marketing and sales, machine learning is used for lead generation, data analysis, online search, and search engine optimization (SEO). For example, many marketers use machine learning algorithms to connect with users who leave products in their shopping carts or exit the website. Additionally, machine learning is used for recommendation engines, such as those on Amazon, Netflix, and StitchFix, recommending based on users’ tastes, browsing history, and cart history.
Optimization: In Meituan’s smart delivery system, operational optimization techniques have been applied to achieve intelligent area planning and smart rider scheduling, increasing delivery efficiency. Optimization techniques play an important role in delivery network planning and capacity structure planning, improving rider route efficiency and reducing empty driving rates through multi-objective optimization problems.
Simulation: Computer simulation is widely used in the engineering field for optimizing design solutions, predicting system performance, and assessing risks. For example, in automotive design, computer simulation can evaluate vehicle fuel economy, performance, and safety; in building design, it can predict earthquake resistance and thermal performance.
Rules/Heuristics: Heuristic evaluation is a comprehensive assessment testing strategy used to detect usability issues in product user interfaces. This process is conducted based on a set of pre-determined usability principles (heuristic methods) and relies on several usability experts conducting an in-depth test to ensure the product provides a seamless user experience.
Knowledge Graphs: Knowledge graphs are applied in healthcare for precise diagnosis of diseases. By constructing a knowledge graph that includes various types of cancers, doctors can quickly match patients’ symptoms with known cancer patterns, improving diagnostic accuracy. Furthermore, knowledge graphs are used in financial risk management, allowing banks and financial institutions to visually reveal risk connections hidden within complex financial networks by constructing knowledge graphs covering financial entities and their relationships.
Causal AI: Causal AI is an emerging technology that attempts to understand the causal relationships between variables in data, rather than just correlations. This technology can be applied in medical research, for example, by analyzing large patient datasets to determine the causal relationship between drug effects and side effects.
Neuro-symbolic AI: Neuro-symbolic AI combines the advantages of neural networks and symbolic processing, aiming to enhance machine reasoning and understanding capabilities. For example, the Neural Rule Engine developed by Deep Curiosity transforms symbolic rule knowledge into a knowledge form of neural networks, supplemented by corresponding computations to improve the recall and precision of rules.
First-principles AI: First-principles AI is a method that starts from fundamental truths and rules to solve problems through logical reasoning. This technology can be applied in materials science to predict the properties and behaviors of new materials from the ground up using basic physical principles.
05 Combining AIGC Models with Other AI Technologies
Three key points:
1. AI technologies are not mutually exclusive; they can often be combined in a way that provides better accuracy, transparency, and performance while also reducing costs and data requirements.
2. Combining AIGC models with other AI technologies is particularly powerful.
3. The potential combinations of AI technologies are endless. Powerful combinations and use cases include:
Combining non-generative machine learning with AIGC models: for segmentation and classification, synthetic data generation, and computer vision
-
Combining optimization/search with AIGC models: for enterprise search
-
Combining simulation with AIGC models: for simulation acceleration
-
Combining graphs with AIGC models: for knowledge management and retrieval-enhanced generation
-
Combining rule-based systems with AIGC models: for chatbots, robo-advisors, and specialized natural language generation.
06 Knowing Three Things to Avoid Going in Circles
1. The hype around AIGC may lead to its use in places where it is not suitable, increasing project complexity and failure risks.
2. Overemphasis on AIGC may lead to overlooking broader alternatives and more mature AI technologies that are better suited for most potential AI use cases.
3. Strive to combine AI technologies to create more powerful systems where different technologies can compensate for each other’s weaknesses.
In the AI era, skills matter more than degrees, and certificates matter more than education!
The Ministry of Human Resources and Social Security is launching AIGC certification, welcome to join me in building differentiated advantages in the AI era.
Extended Reading
-
AI Skills Are the New Favorite! Survival Guide in the Era of Generative AI
-
“An Anxious Generation”: How Digital Technology Profoundly Affects the Mental Health of Generation Z, and 5 Practical Suggestions
-
[By Percentage] The Fastest Growing and Declining Jobs from 2025 to 2030
-
The 15 Jobs with the Largest Net Growth and Decline in Absolute Numbers from 2025 to 2030
-
World Economic Forum: By 2030, 170 million jobs will be created, 92 million jobs will be replaced, and 1 billion jobs will face transformation
Copyright Notice:“Lu Shan Says AI Improvement” The articles on the personal public account are all original. Reprinting, excerpting, or copying any usage without my permission is prohibited. For reprinting, quoting, or other intentions, please apply in advance through the backend of this public account and obtain authorization. Some texts and images in original articles are sourced from the internet; for any issues, please contact this public account.