Understanding AI Alignment — the science of making AI systems do what we actually want — explained for everyone.
تێگەیشتن لە هاوڕێککردنی زیرەکیی دەستکرد — زانستی وا کردنی سیستەمەکان کە ئەوە بکەن کە بەڕاستی دەمانەوێت.
01 / ALIGNMENT
هاوڕێککردنی زیرەکیی دەستکرد چییە؟
Alignment is the challenge of making AI systems pursue the goals we intend — not just the goals we literally specify. A system optimizing the wrong objective can cause harm even without any bad intent. As models grow more capable, getting this right becomes one of the most important technical problems of our time.
02 / INTERPRETABILITY
سەیرکردنی ناوەوەی سندوقە ڕەشەکە
Interpretability research tries to understand why a neural network produces the answers it does. If we can read a model's internal reasoning, we can catch deception, bias, and errors before they matter — like an MRI for artificial minds.
03 / EVALUATION
تاقیکردنەوەی مۆدێلە پێشکەوتووەکان
Before deployment, frontier models are stress-tested for dangerous capabilities and failure modes — red-teaming, benchmark evals, and safety cases that must hold up under adversarial pressure.
04 / GOVERNANCE
سیاسەت و گەشەپێدانی بەرپرسانە
Technical safety needs institutional support: scaling policies, model audits, and international coordination so that safety keeps pace with capability everywhere — including our region.
"The goal is not to slow AI down.
It's to make sure it goes well."
ئامانج خاوکردنەوەی زیرەکیی دەستکرد نییە — دڵنیابوونە لەوەی بە باشی بەڕێوە دەچێت.
Start with our beginner-friendly introduction — no math, no jargon, available in Kurdish.
Start the Intro Course