Ben Torben-Nielsen, PhD, MBA on LinkedIn: #ai #education #epfl
Could ChatGPT get an engineering degree? According to EPFL, it could.
Recently, I just read a fascinating study from EPFL. Without any optimization, GPT-4 correctly answered 65.8% of all exam questions across 50 engineering courses. With proper prompting, this number increased to 85.1%.
Future Student Skills
These findings raise (at least) two fundamental questions.
First, what skills should we teach future students? Think of calculators that changed how we teach math, but did not remove the need to understand basic concepts. AI will reshape engineering education the same way. We must decide which core skills matter most.
Testing Students
Second, how do we test students when AI can solve standard problems? Simple exams may not work anymore. Instead, we might need tests that combine many concepts and require deeper thinking.
Education and AI Integration
To answer your questions. Again in my class I teach how to use GenAI (chatGPT , copilot , and Gemini to help coding in Google Colab). We expect them to use it as I use it for research. The evaluation is similar to a real job.
Balance in Education
All education and learning, simplified and accessible in one place—ChatGPT. Ben Torben-Nielsen, PhD, MBA
Human vs. AI Learning
If a student were given access to all the books and the Internet that ChatGPT is trained on, with ChatGPT’s computational speed scaled to that of a human, the student would likely outperform the AI. Human learning involves synthesizing, adapting, and applying knowledge critically in dynamic situations. With no time limit, the students can improve, whereas ChatGPT cannot unless a human behind the curtain can improve it further.
Exams and Projects
Assignments and projects allow for research, collaboration, and creativity, making them ideal for integrating AI. However, exams still hold unique value. Exams challenge real-time understanding, critical thinking, and problem-solving, fostering essential skills like reasoning and adaptability that AI cannot yet replicate.
Transformative Shift in Education
The integration of knowledge-generation tools should drive a transformative shift in education, not just minor adjustments. While educational practices have varied across nations and academic levels, the focus must shift from knowledge retention to fostering critical thinking and skills-based learning. This isn’t about making exams harder, but about rethinking curricula, modernizing teaching with evidence-based practices, and designing tasks that promote critical thinking and hands-on learning while considering individual learning differences. To achieve higher-quality work, AI should be embraced as a collaborative tool.
Data & AI Strategist & Enterprise Architect | Data & AI-driven transformation | 25+ Years Experience: Mobile +47 900 22 715
The Learning Dilemma
The Learning Dilemma: studies have unveiled a striking paradox in programming education: While AI-assisted coding tools offer the fastest solution to coding problems, they result in the lowest retention of knowledge. In contrast, traditional methods like using search engines, though slower, lead to significantly better recollection and learning outcomes.
So we are among to shifting skills from code crafters to "AI whisperers" without any understanding more than be a whisper....?
