[key points]

Key points

• •The article discusses the concept of "alignment" in artificial intelligence, which aims to ensure that machines behave according to human expectations in terms of reliability, safety, and ethics.
• •Robotics engineers have found that defining instructions can lead to unexpected behaviors, as natural language commands contain implicit provisions that are difficult for computer systems to translate.
• •With the democratization of conversational agents like ChatGPT, the issue of aligning machines with human values has become a public debate, prompting AI companies to invest in training phases to reduce inappropriate or immoral outputs.