At the end of 2022, Large Language Models (LLMs) burst onto the scene. ChatGPT, along with its derivatives, such as the Junmou system, is gradually becoming an indispensable resource for every learner. Particularly within university campuses, a growing number of students have started utilizing these LLMs to assist in knowledge acquisition and assignment completion. The potency of this new tool lies in its capability to provide elaborate answers by simply posing a query. Moreover, their user-friendly design includes features like “regenerate” button, allowing users to easily request alternative responses if the initial answer doesn’t fully meet their needs. This level of accessibility and adaptability positions LLMs as a formidable ally in the academic journey, and each learner seems to have acquired an abundance of resources overnight.
Napoleon Bonaparte once declared, “Every French soldier carries a marshal’s baton in his knapsack.” Now this sentiment finds a new form: every learner wields their own marshal’s baton through their laptop, assuming the role of a general commanding LLMs. Despite the formidable capabilities of these AI-driven armies, they are not without a critical weakness — they cannot make confident judgments regarding their output or determine the accuracy of their responses. LLMs, at their core, are advanced natural language processing tools. Their answers are inherently confined by the extent and caliber of their training data, as well as the underlying algorithm. The mighty army of LLMs, while providing learners with significant services, also presents learners with great challenges: should learners use it? How can we ascertain the accuracy of its responses? And how to guide LLMs to provide better answers?
Change of Mindset: From Resistance to Engagement
The key for learners to move from their past role as “soldiers” to their current role as “generals” is whether they recognize this change as an inevitable trend and accept that the LLM is unavoidable for today’s learners. The presence of LLMs is constant and independent of individual usage; they don’t cease to exist simply because someone chooses not to use them, and they are not exclusively operational for those who do. Consequently, several renowned international universities have shifted their stance toward LLMs, transitioning from initial resistance to embracing acceptance. Learners need to embrace their role as “generals”, adopting a mindset of positive engagement rather than negative resistance.
Core Competencies Highlighted: Critical Thinking
The transformation in the role of learners has brought about a shift in the essential skills they must acquire. For the “soldier,” tactical prowess and the ability to independently grasp every detail of knowledge are paramount. However, for the “general,” the key skill lies in strategic learning — the ability to oversee the entire situation and evaluate the accuracy of answers provided by the “soldiers.” This entails a new challenge for learners: independently assessing and verifying the outputs generated by LLMs from their extensive databases. Consequently, critical thinking skills have gained prominence, becoming a fundamental requirement for learners. This involves distinguishing between opinions and facts, assessing the logical coherence of viewpoints, scrutinizing the sufficiency of the evidence supporting an opinion, and identifying and reflecting upon the underlying assumptions of opinions and arguments. All these elements form the essence of critical thinking.
New Knowledge Structure: Broad but Precise
To command better answers from large language models, mere critical thinking to judge the accuracy of knowledge is not sufficient; a robust knowledge base is also essential. Traditional education emphasizes a “narrow and deep” approach to knowledge acquisition, where learners are expected to gain extensive and detailed expertise in a specific domain, ultimately becoming specialists in particular fields. However, LLMs necessitate learners to grasp core principles and fundamental aspects across a wide spectrum of knowledge domains, constructing a comprehensive and interconnected knowledge network. This represents a “broad but precise” knowledge structure. While the “narrow and deep” knowledge can be substituted by the LLMs to a certain extent, the “broad but precise” knowledge exceeds the current capabilities of such models. Moreover, the learner, as “general”, can use this “broad but precise” knowledge to command the LLMs to produce better answers.
In summary, while LLMs significantly empower learners, they also present new challenges. First and foremost, learners must adjust their mindset, actively embracing and adapting to this novel paradigm. Secondly, learners must place greater emphasis on cultivating critical thinking to accurately discern the validity of knowledge. Furthermore, learners’ knowledge structures must transition from being “deep and narrow” to “broad but precise”, enabling effective guidance of LLMs to produce higher-quality responses, ultimately achieving the transformation from being a “soldier” to becoming a “general.”
