Understanding how AI can affect the economy and society – and the education system that prepares students for both – requires an understanding of the capabilities of this technology and their development trajectory. Moreover, AI capabilities need to be compared to human skills to understand where AI can replace humans and where it can complement them. This knowledge base will help predict which tasks AI may automate and, consequently, how AI may shift the demand for skills. Policy makers can use this information to reshape education systems in accordance with future skills needs and to develop tailored labour-market policies.
Artificial intelligence and education and skills
As AI rapidly advances, it is crucial to understand how education will be affected. Important questions to consider include: How can AI be compared to humans? How do AI systems perform tasks from various capability domains such as language, reasoning, sensorimotor or social interaction domains? What are the implications for education and training?
Key messages
As AI rapidly advances, it's becoming evident that it is starting to outpace humans in critical areas such as reading, mathematics and scientific reasoning. This prompts us to reconsider our educational approach. We must determine which skills to prioritise, which to phase out, and where to place greater emphasis in an AI-influenced world. We need to anticipate how learning methods and teaching practices will evolve. At the same time, more profound questions about the overall goals of education are emerging as the cognitive, physical and social capabilities of AI continue to rise.
Countries should encourage collaborative research on effective and equitable use of generative AI in the teaching and learning process to inform forward-looking guidance and dedicated training programmes. Monitoring impact and sharing (international) best practices across researchers, developers, and education stakeholders will help cast light on the multiple benefits of generative AI, as well as on its limitations, allowing innovation and improvements while mitigating pitfalls.
Context
How do GPT and student performance in PISA compare?
The performance of GPT – the AI system behind OpenAI Chatbot ChatGPT – on reading and science is at a higher level than students. GPT-3.5 (released in November 2022) can solve 73% of the reading test questions and 66% of the science questions, while GPT-4 (a more powerful version released in March 2023) scores at 85% and 84%, respectively. In contrast, GPT-3.5 and GPT-4 mathematical capability proved to be still below that of students. Moreover, GPT-4, released only a couple of months after its predecessor, performs at a substantially higher level for each capability .
GPT and student performance on PISA core domain
Evolution of a written task: exploring human-AI synergies
With AI’s rapid growth, tasks humans perform today are likely to change in the future. In consequence, exploring the way humans use, rely on or collaborate with AIs is necessary to adjust the way we should rethink education systems in light of AI capabilities. The recent advent of ChatGPT gives us a concrete example on how written tasks can evolve with AI and its increasing capabilities.
