Can ChatGPT Pass a Top-Tier Engineering Coursework?

Описание: One of the biggest questions in education today: Can AI solve complex engineering coursework? To find out, I ran an experiment using the actual assignment from my master-level module Applied Advanced Optimization at Imperial College London.

This is highly technical coursework involving coding, mathematics, pattern recognition, and open-ended analysis, designed to stretch top students over two months. I gave this exact assignment to Meta AI, Claude, ChatGPT-4, and Gemini 2.5 Pro. Did they manage?

💡 Could a student with zero understanding really copy-paste their way to top marks?
⚠️ What does this mean for academic integrity and the future of assessment in engineering?

🔬 In this video:

I share how each AI performed
I reveal how these AI-generated submissions ranked against real students
I discuss the troubling implications of this experiment

📄 Full coursework, AI outputs, and source code available here:
Gemini coursework: https://colab.research.google.com/dri...
Gemini chat: https://aistudio.google.com/app/promp...
ChatGPT coursework: https://colab.research.google.com/dri...
ChatGPT chat: https://chatgpt.com/share/681f660e-53...
Meta AI coursework: https://colab.research.google.com/dri...
Clude AI coursework: https://colab.research.google.com/dri...
Coursework questions: https://colab.research.google.com/dri...

This is a must-watch for educators, students, and anyone concerned about the role of AI in education.

*CHAPTERS:*
00:00 Introduction
01:13 Description of the coursework
02:46 The AI challenge
03:25 Ground rules of the experiment
04:49 Rule 4: markers do not know!
05:10 The worst AI
06:37 3rd place
07:31 2nd place
09:31 The winner!
11:32 Implications and reflection
12:38 Solution 1: the race
13:34 Solution 2: death?
14:42 Solution 3: AI-enhancing