Airlines are adopting artificial intelligence (AI)-driven automation for predictive maintenance and customer service.

AI is a technology that is evolving at an unprecedented speed, especially since the rise of modern Generative AI applications (democratized to the public with ChatGPT), making it difficult to keep track of its state of the art (SoTA). Generative AI currently receives much of the hype, potentially creating the misperception that it represents the sole SoTA. However, it is also crucial to consider more “familiar” and well-proven techniques. Most of the value delivered by current industrialized projects comes from classical AI techniques, sometimes mixed with Generative AI. Generative AI focuses on creating or using models that can autonomously produce new content, such as images, text, or videos, based on patterns learned from existing examples. Generative AI-based chatbots often interact with users through natural language prompts and are mainly based on largelanguage models (LLMs), which are a type of machine learning (ML) model trained on a vast amount of data. More traditional techniques include ML time series processing (e.g. for predictive maintenance and anomaly detection), Natural Language Processing (e.g. for automated speech-to-text recognition), Computer Vision (e.g. for vision-based landing and defect detection from images), and Reinforcement Learning (e.g. teaching robots to perform manufacturing and assembly tasks). Finally, Quantum Computing, built upon established AI techniques and relying on quantum computational architectures, offers a paradigm shift in problem-solving by achieving significantly faster processing times for computationally intensive tasks, such as optimization, simulation, and ML.For specific tasks, such as object detection from images, algorithms powered by deep learning have reached SoTA performance surpassing humans. Methods and tools for explainability and robustness of AI—especially as models become larger—remain an evolving area of research. a) The stakes for AI are high, with business value foreseen in streams such as: b) AI for operational efficiency, by introducing new ways of working and saving time compared to traditional approaches; c) AI embedded in products for distinctive and disruptive functionalities, such as enhancing safety beyond standards; d) AI for enhancing market reach and brand image; e) AI for developing new revenue models. Understanding the multifaceted nature of AI, including both its potential benefits and inherent risks, is crucial for navigating this evolving technological landscape responsibly. The top risks to be properly managed include the loss of competencies and critical thinking (when AI is used to gain operational efficiency), the lack of model explainability (especially for embedded applications), and the improper or unethical use of AI technology. Other threats concern the risk of losing track of the source of AI-generated information, job displacement, cybersecurity vulnerabilities, data privacy concerns (especially for ML), or even the environmental impact of generative AI (model training and hosting are highly power-consuming).