0:00 / 0:00
research
The Skills AI Can’t Replace 🧠⚙️
MIT CSAIL16 June 2026Watch on YouTube
Part of series
Ep. 10 · Waarom AI liegt
Onderzoek naar de oorzaken van AI-hallucinaties en waarom gebruikers niet kunnen vertrouwen op de betrouwbaarheid van taalmodellen.
View the seriesDescription
AI is getting insanely good, but mostly at things that are easy to check. So what happens to the skills that aren’t instantly testable? MIT CSAIL Associate Director Professor Armando Solar-Lezama explores why future learning needs to shift toward long-term thinking, design decisions, and experience-driven skills, especially in fields like software development where mistakes don’t show up until much later. #AI #FutureOfWork #Education #MachineLearning #SoftwareEngineering #CriticalThinking #TechTrends #Learning
What you'll learn
- AI excels at tasks that are easily verifiable, but struggles with skills that can only be evaluated over the long term
- Long-term thinking and anticipating future consequences are core capabilities that AI cannot replace in software development
- Design decisions and experience-driven learning remain essential because code errors often don't surface until much later
- Education must shift toward building human capacities that aren't instantly testable, rather than focusing on AI-replaceable skills
Frequently asked questions
What types of tasks does AI perform well?
According to Solar-Lezama, AI excels at tasks that are easily checkable, where results can be verified immediately.
Why is long-term thinking important in software development?
Errors and consequences of design decisions in software often only become apparent much later, requiring developers to think about future impact.
How should education adapt to the rise of AI?
Education should focus more on developing human skills that aren't instantly testable, such as experience-driven learning and critical thinking.
What are examples of skills AI cannot replace?
Design decisions, long-term planning, and learning from experience are skills AI cannot replace, especially in complex fields like software development.