(The copyright belongs to Gartner and is for learning only)

core trends

• current Situation and Future of generative AI: GenAI has exceeded the expected expansion period, but the hype surrounding it still exists. In 2024, more value will come from projects based on other AI technologies. Whether it is independent application or combination with generative AI, these projects will benefit from standardized processes to promote implementation. The future system architecture should be based on composite AI technology, which means that AI leaders should integrate innovative methods at all stages of the technology maturity curve to maximize benefits.
• The rise of AI Engineering and Knowledge Graph: AI Engineering and Knowledge Graphs have made remarkable progress on the technology maturity curve in 2024, highlighting the demand for robust methods when processing AI models on a large scale. AI engineering is the basis for the delivery of enterprise-level AI solutions. The Knowledge Graph provides reliable logic and interpretable reasoning, which is in contrast to the deep learning technology used by generative AI.
• Wide application of composite AI: Composite AI will become the standard method for developing AI systems within two years and will be widely used. This method combines a variety of AI technologies to enhance learning efficiency and expand the scope of knowledge representation, thus effectively solving a wider range of business problems.
• The future of autonomous systems: autonomous Systems are becoming an important trend because they can achieve business adaptability, flexibility and agility that traditional AI technologies cannot achieve independently. The learning ability of these systems is especially valuable when the operating environment is unpredictable and real-time monitoring and control are not feasible.

Suggestion

1. Develop an AI governance framework:

in view of the increasing importance of responsible AI and AI TRISM, enterprises should establish a comprehensive AI governance framework as soon as possible to ensure that AI applications comply with ethical and regulatory requirements.

2. Investment edge AI and composite AI:

enterprises should pay attention to edge AI and Composite AI technologies to improve the performance, adaptability and privacy protection of AI systems.

3. Reshape the data strategy:

the importance of Synthetic Data is increasing. It is recommended that enterprises incorporate it into their Data strategies, especially in scenarios where privacy is sensitive or Data is scarce.

4. Strengthen AI engineering capabilities:

enterprises should invest in AI engineering and ModelOps related skills to establish an end-to-end AI lifecycle management process to prepare for large-scale AI deployment.

5. Explore vertical applications:

as intelligent applications enter a period of steady rise and recovery, enterprises should actively explore deep AI applications in their vertical fields, such as computer vision and autonomous driving.

Gartner AI technology maturity curve in 2024

technology germination (Innovation Trigger)

adaptive Systems

adaptive system refers to a system that can automatically manage and optimize its own operations. These systems use artificial intelligence and machine learning technologies to automatically adjust their behaviors to adapt to environmental changes without human intervention. They are usually used in complex IT infrastructure management, network security, and intelligent transportation systems. The adaptive system has the following key features:

• Autonomy: able to make decisions and perform tasks independently without external assistance.

• Learning ability: able to adjust its behavior and internal operations according to experience, changing conditions or goals.

Agency: have the perception of their own internal state and goals, guide their learning methods and behaviors, and enable them to act independently.

Quantum AI

• ultra-high parallelism: it uses the superposition and entanglement characteristics of quantum bits to achieve faster computing speed than classical computing.
• Ability to solve complex problems: able to efficiently handle complex optimization and simulation problems that cannot be solved by traditional computing.

Representative manufacturer: Google's quantum computing platform is dedicated to developing quantum algorithms to improve machine learning performance.

First-Principles AI (FPAI)

• root cause analysis: focus on the basic principles of the problem rather than surface phenomena.
• Model generalization ability: it can apply knowledge more effectively in untrained scenarios.

Representative Vendor: DeepMind promotes the application of AI in scientific research through models based on physical principles.

Artificial intelligence (AI)

• physical interaction capability: able to interact directly with the real world.
• Environmental adaptability: adjust behaviors and decisions according to environmental changes.

Representative manufacturer: Boston Dynamics's robot system demonstrates the application capability of AI in complex environments.

Multi-Agent Systems

• collaboration capability: multiple agents can work together to achieve common goals.
• Independent decision-making: each intelligence can make decisions and take actions independently.

Representative manufacturer: IBM's multi-agent platform is used to optimize supply chain management.

AI Simulation

• efficiency: use AI technology to improve the efficiency and accuracy of simulation.
• Diversity: supports multiple simulation models to adapt to different application scenarios.

Representative Vendor: AnyLogic provides powerful simulation tools to support modeling requirements in various industries.

Causal AI

• causal reasoning ability: able to identify and use causal relationships for prediction.
• Decision Optimization: provide more effective action suggestions based on causality.

Representative manufacturers: representative companies of Causal AI, such as cabeautens, focus on the application of Causal reasoning technology.

Artificial intelligence Data applicability (AI-Ready Data)

• data governance capabilities: Ensure Data compliance and effectiveness during use.
• Use case adaptation: Allows you to adjust data management policies based on different AI application scenarios.

Representative Vendor: Databricks provides a comprehensive data management platform to support the applicability of AI data.

Decision Intelligence

• Comprehensive decision support: provides decision suggestions based on multiple data sources and analysis methods.
• Feedback Optimization: improve the decision-making process through continuous feedback mechanism.

Representative Vendor: Microsoft's Azure decision intelligence platform integrates a variety of data analysis and AI tools.

Neural symbol artificial intelligence (Neuro-Symbolic AI)

• fusion Ability: combines symbolic reasoning with the learning ability of neural networks.
• Enhanced reasoning: provides stronger reasoning and generalization capabilities in complex tasks.

Representative manufacturer: Google's DeepMind is in the leading position in the research of neural symbol AI.

Composite AI

• technology integration capability: combine different AI technologies to enhance system capabilities.
• Flexibility: it can be quickly adjusted and optimized according to different requirements.

Representative Vendor: SAP's AI platform supports enterprise digital transformation through composite AI technology.

Artificial General Intelligence

• extensive adaptability: able to perform well in a variety of tasks and environments.
• Autonomous learning ability: ability to learn from experience and self-improvement.

Representative manufacturers: Although AGI is still in the research stage, the exploration of OpenAI, DeepMind and other companies in this field has attracted much attention.

Sovereign AI

• autonomy: The country can independently formulate and implement AI strategies.
• Resource Utilization: effectively utilize domestic data and technical resources to promote development.

Representative manufacturers: governments cooperate with local technology companies to promote the development of sovereign AI.

Peak of Inflated Expectations

AI Trust, risk, and security management (AI TRiSM)

AI TRiSM is a framework proposed by Gartner to help organizations identify and mitigate reliability, security, and trust risks associated with artificial intelligence. The framework consists of four pillars: interpretability and model monitoring, model operation, AI application security and privacy, ensuring the transparency and compliance of AI systems. AI TRiSM ensures the governance, credibility, fairness, reliability, robustness, effectiveness, and data protection of artificial intelligence. Its key features include:

• transparency: ensure that the operation process of models and applications can be understood and reviewed.

• Exception detection: identify and handle content exceptions in a timely manner to maintain system security and compliance.

• Data protection: ensures data security and privacy protection during use.

Representative Vendor: Lasso Security provides a comprehensive Security solution in the AI TRiSM field, focusing on content exception detection and privacy protection.

Prompt Engineering

• input optimization: guide the model to generate desired output by precisely designing input prompts.
• Example Guide: use the example to further guide the model generation process.

Representative Vendor: OpenAI is widely used in prompt engineering, especially in its GPT series models.

Responsible AI

• Ethical considerations: integrating ethical and social responsibilities into the design and application of AI systems.
• Bias mitigation: take measures to reduce bias in the algorithm and ensure fairness.

Representative manufacturers: IBM's practices and policies in responsible artificial intelligence are widely recognized.

AI Engineering

• lifecycle management: comprehensive governance from model building to deployment and monitoring.
• Scalability: ensure that the AI system can be expanded quickly as needed.

Representative Vendor: Microsoft provides powerful tools and platforms in the field of AI engineering to support enterprises to deploy AI solutions on a large scale.

Edge AI

• real-time processing: data is processed at the data generation site to reduce latency.
• Local inference: supports AI inference on edge devices to reduce dependence on the cloud.

Representative manufacturer: NVIDIA is in the leading position in edge AI hardware and software solutions.

Foundation Models

• self-supervised learning: pre-training through self-supervised mode to improve the generalization ability of the model.
• Multi-modal adaptation: able to process multiple types of data and tasks.

Representative Vendor: The GPT series of OpenAI and the Llama model of Meta are typical representatives of the basic model.

Synthetic Data

• data anonymization: protects the privacy of real data by generating composite data.
• Enhanced model training: provides additional training data sources when data is scarce.

Representative Vendor: Hazy focuses on generating synthetic data and provides data solutions for various applications.

ModelOps

• end-to-end governance: comprehensive management of the model lifecycle.
• Dynamic update: continuously optimizes model performance based on real-time feedback.

Representative Vendor: DataRobot provides comprehensive solutions for model operation and supports enterprise AI Model Management.

AI

• content creation capability: it can generate multiple types of content to meet different needs.
• High authenticity: the generated content has high authenticity and relevance.

Representative Vendor: OpenAI's DALL-E and ChatGPT have important influence in the field of generative AI.

Bubble burst Trough

Neuromorphic Computing

• high efficiency: high performance and low power consumption computing are realized by simulating the structure of biological neural networks.
• Adaptive learning: capable of dynamic learning and adjusting computing strategies based on environmental changes.

Representative manufacturer: Intel's Loihi chip is an important example of neural morphology computing. It uses the peak neural network (SNNs) architecture to significantly improve energy efficiency.

Smart Robots

• Independent decision-making ability: able to make independent decisions in complex environments.
• Task adaptability: adjust the behavior according to different tasks and environmental conditions.

Representative Vendor: Boston Dynamics's robot system demonstrates the ability to perform multiple tasks in complex environments.

Cloud AI Services

• convenience: quickly build and deploy AI models through the cloud platform to reduce the technical threshold.
• Scalability: you can flexibly adjust resources based on your needs and support large-scale applications.

Representative Vendor: Amazon Web Services (AWS) provides comprehensive cloud artificial intelligence Services, including machine learning and data analysis tools.

Steadily climbing the recovery period (Slope of Enlightenment)

Autonomous Vehicles

• autonomous navigation capability: able to identify and handle traffic conditions in complex environments.
• Multi-sensor fusion: provides 360-degree environmental perception through the cooperative work of various sensors.

Representative manufacturer: Tesla's Autopilot system shows the autonomous driving ability under various driving conditions.

Knowledge Graphs

• relationship Modeling: A complex information network is represented by nodes (entities) and edges (relationships).
• Inference capability: supports inference of relationships between entities to improve the effectiveness of information retrieval and understanding.

Representative Vendor: Google's Knowledge Map provides its search engine with powerful information organization and reasoning capabilities.

Intelligent Applications

• personalized Experience: provides customized services based on user behaviors and preferences.
• Adaptive Capability: it can automatically adjust response policies in different scenarios.

Representative Vendor: the intelligent application of Salesforce improves the efficiency and effectiveness of customer relationship management through data analysis and machine learning.

Production maturity (Plateau of Productivity)

Computer Vision

• image recognition capability: able to recognize and classify objects, scenes and activities in images.
• Real-time processing: supports real-time analysis of video streams and is suitable for applications in dynamic environments.
• Deep learning integration: use deep learning algorithms to improve the accuracy and efficiency of image processing.

Representative manufacturer: NVIDIA provides powerful hardware and software solutions in the field of computer vision, supporting various application scenarios, including autonomous driving and intelligent monitoring.

Gartner AI technology maturity curve in 2023

technology germination (Innovation Trigger)

Automatic Systems

Automation system refers to a system realized by technology, program, robot or process, which can complete tasks with minimal manpower input. These systems usually integrate sensors, controllers and actuators to realize self-management and optimization, and are widely used in industry, commerce and daily life.

First-Principles AI

artificial intelligence based on first principles is a method to build AI models based on basic principles, emphasizing the fundamental understanding of problems rather than relying on experience or existing models. This approach helps to develop more innovative and effective AI solutions.

Multi-Agent Systems

multi-agent system is a distributed system composed of multiple agents, which can perceive, learn and act independently to achieve their own and common goals. They perform well in complex environments and are suitable for collaborative tasks and dynamic systems.

Neural symbol artificial intelligence (Neuro-Symbolic AI)

neural symbol artificial intelligence combines the deep learning ability of neural network and logical reasoning ability of symbolic reasoning, aiming at improving the understanding and reasoning ability of AI system. This method can better deal with complex cognitive tasks.

Causal AI

causal artificial intelligence focuses on identifying and understanding causal relationships in data, surpassing traditional correlation analysis. Through causal reasoning, causal AI can provide deeper explanation and optimization ability, which is suitable for decision analysis and policy formulation.

AI Simulation

artificial intelligence simulation is to simulate systems and processes in the real world through computer models for analysis and prediction. This technique can be used to train machine learning models, test the effectiveness of algorithms, and make decision support.

AI Engineering

AI engineering applies software engineering principles to the development and deployment of artificial intelligence systems, covering the whole process of model construction, testing, deployment and maintenance to ensure the reliability and scalability of AI solutions.

Data-Centric AI

AI emphasizes the core role of data in AI systems and focuses on data quality, availability, and management to optimize model performance and effectiveness. With the data-driven approach, AI systems can better adapt to different application scenarios.

Composite AI

composite Artificial Intelligence integrates multiple AI technologies to create more complex and flexible intelligent systems. By combining different technologies, composite AI can solve complex problems more effectively and provide more comprehensive solutions.

Operational AI Systems

operational artificial intelligence system refers to an AI system that can make real-time decisions and optimize in actual operation. These systems usually integrate machine learning and automation technologies to improve the efficiency and flexibility of business processes.

AI Trust, risk, and security management (AI TRiSM)

AI TRiSM is a framework designed to help organizations identify and mitigate Trust, risks, and security issues related to artificial intelligence. It includes interpretability, model monitoring and compliance to ensure the transparency and compliance of AI systems.

Decision Intelligence

decision intelligence combines data analysis, artificial intelligence and human judgment to optimize the decision-making process. It emphasizes how to use data to make effective decisions and is widely used in commercial, medical and government fields.

Artificial General Intelligence

artificial general intelligence (AGI) refers to an intelligent system that can understand, learn and apply knowledge to solve various problems, similar to human intelligence. AGI's goal is to achieve human-level intelligence in a variety of tasks and fields.

Peak of Inflated Expectations

Prompt Engineering

Prompt engineering refers to the process of designing and optimizing input prompts to improve the output quality and relevance of generative AI models (such as GPT-3). With accurate prompts, users can guide AI to generate more expected content, which is widely used in natural language processing and generation tasks.

Neuromorphic Computing

neural morphology computing is a computing technology that imitates the structure and function of the human brain, aiming at improving computing efficiency and processing speed. It optimizes the execution of AI algorithms by simulating the working principles of neurons and synapses.

Responsible AI

responsible artificial intelligence emphasizes ethical principles and social responsibilities when developing and applying AI technology. This includes ensuring fairness, transparency, and interpretability of the algorithm to avoid prejudice and improper use.

Smart Robots

intelligent robots refer to robots that can sense, make decisions and perform tasks independently. They are usually combined with machine learning and sensor technology and are widely used in manufacturing, medical treatment and service industries.

Foundation Models

A basic model is a deep learning model that is pre-trained on large-scale datasets and can be adapted to various tasks through fine tuning. This kind of model shows strong transfer learning ability in natural language processing, computer vision and other fields.

AI

generative artificial intelligence refers to AI technology that can generate new content (such as text, images or audio), and uses deep learning models to create highly authentic content, it is widely used in fields such as artistic creation and content generation.

Synthetic Data

synthetic data is artificial data generated by algorithms to simulate the characteristics of real data. It is usually used to train AI models, especially in the case of privacy protection and data scarcity.

Bubble burst Trough

ModelOps

ModelOps is a practice that focuses on the governance and lifecycle management of artificial intelligence and decision models. It is designed to ensure a smooth transition from development to production, covering the deployment, monitoring, update, and compliance of models. ModelOps is similar to DevOps, but focuses on data analysis and AI model management to help enterprises implement continuous model delivery and efficient operation.

Edge AI

edge artificial intelligence refers to AI technology that processes and analyzes data at locations near data sources (such as devices or sensors). This method reduces latency and improves the efficiency of real-time data processing. It is suitable for smart home, autonomous driving, industrial Internet of Things and other application scenarios.

Knowledge Graphs

knowledge Graph is a data structure used to represent entities and their relationships, which can help AI systems understand and infer complex information. They play an important role in search engines, recommend systems, and natural language processing. They enhance data availability and understanding by providing semantic contexts.

AI Maker and Teaching Kits

AI production and teaching toolkit is a tool and resource designed for education and developers to help users learn and build AI applications. These toolkits usually include programming environments, sample projects and teaching materials, and are suitable for schools, training institutions and individual developers to promote the popularization and application of AI technology.

Autonomous Vehicles

self-driving vehicles refer to vehicles that can navigate and drive independently without human intervention. These vehicles use sensors, AI algorithms and real-time data processing technologies to improve traffic safety and efficiency and are gradually becoming an important part of future traffic.

Steadily climbing the recovery period (Slope of Enlightenment)

Intelligent Applications

intelligent application refers to an application that integrates AI technology and can provide personalized and intelligent user experience. These applications usually use machine learning and data analysis to optimize user interaction and services, and are widely used in business, medical and customer service fields.

Cloud AI Services

AI is an AI tool and service provided by the cloud computing platform, allowing enterprises to quickly deploy and expand AI solutions. These services usually include machine learning models, data analysis, and natural language processing, lowering the threshold for AI technology.

Data Labeling and Annotation

data tagging and annotation is a process of classifying and annotating data for training machine learning models. High quality data tagging is the key to ensure the accuracy and effectiveness of AI models and is widely used in computer vision, natural language processing and other fields.

Computer Vision

computer vision is a technology that enables computers to understand and interpret visual information (such as images and videos). It plays an important role in the fields of automatic driving, medical imaging and security monitoring, and promotes the development of intelligent equipment.

Production maturity (Plateau of Productivity)

<none>

---[This article is finished]] ---

this article is reprinted from: Andy730