香港科技大學副教授黃岳永 erwin ：「我們更加想知你如何使用(ChatGPT)，不是量度結果而是量度過程。將來老實說，電腦1分鐘便可完成，即使我給你30分鐘，你都是1分鐘完成。但當你如果1分鐘便可以找到某些資料，你可否用29分鐘的時間，將這些資料變成一些啟發，做一些更多好的表達方式。圖像、啟發，各樣做得更好，那便把自己的分數、能力、理解更為提升。」
GPT Challenge and Design 03_23 details, w reflections..
Gamma.app ai generated
Interview Reflections: The experiment was fascinating and provided many insights I would not have gleaned otherwise. Upon realizing ChatGPT was unavailable, I felt utterly unfamiliar and strange. Over the past months, I have grown so accustomed to ChatGPT that without it, I feared losing my ability to gather information quickly from other sources like Google, as I have not been practicing as much.
However, I firmly believe artificial intelligence will not replace us but instead enhance the capabilities of those who understand it best. ChatGPT is an incredibly powerful tool, but how it is applied is equally important. Repetitive, mundane tasks and rapid generation of outlines and details benefit from its use. But for creative thinking, human intellect is vital.
The experiment demonstrated that empowering the human user with ChatGPT is more critical than allowing ChatGPT to empower the user. ChatGPT alone often yields mediocre results. But with proper guidance, humans and AI collaborating can achieve superior solutions unobtainable by either alone. ChatGPT may supply a B-level response, but under human direction, together they co-create A+ answers.
In short, the key takeaways were: 1) We quickly become dependent on AI tools like ChatGPT and lose other skills if we do not practice them. 2) Human judgment remains essential for high-quality creative work—AI can only assist, not replace, us. 3) How we apply AI tools like ChatGPT is crucial. They should augment human users, not operate independently. 4) With the proper partnership, people and AI can accomplish more than either alone. But human leadership is the key to success. (This reflection is aided by ChatGPT.)
- Some thoughts:
- Reflecting on the activity: GPT is very efficient in terms of structuring things, e.g. giving a presentation flow, syllabus outline, etc. I used to have GPT summarize articles for me so that I could have a grasp of the ideas quickly. To note, the accuracy of GPT is questionable when tasks are very specific, e.g. "name the art style of Yuan Dynasty painter Huang Gongwang" or "provide a criticising argument on the academic paper XXX". GPT would work more on general ideas or facts that are considered "universal knowledge", e.g. "what is quantum physics". Aside from GPT, it is great to have learnt about new AI tools, e.g. Gamma.app, Midjounrney...
- Going forward: The activity of doing a timed research in 30min including output is a very good exercise that sparked me to do the same going forward. In the past week, I listed multiple topic of interest to research on, and I set a 30min period each day for myself to conduct research. This helps with researching my area of interest in an orderly and efficient manner. The help of GPT is also very useful given the time constrain. It helps also with establishing a general idea of the subject - detailed matter maybe not be accurate but it's good enough for this mini "curiosity project".