Chapterبەش 1 · Lectureوانە 1.1

What Is Military AI?ژیریی دەستکردی سەربازی چییە؟

What military AI is and the many jobs it does (most of them are not weapons), what AI can and cannot do, the four reasons armed forces want it, each with its honest limit, dual use, and the map of this course.

چییە و ئەو کارە زۆرانەی دەیکات (زۆربەیان چەک نین)، چی دەتوانێت و چی ناتوانێت، چوار هۆکارەکەی هێزە چەکدارەکان بۆ ویستنی، هەریەکە لەگەڵ سنووری ڕاستەقینەی خۆی، و نەخشەی ئەم کۆرسە.

13 min read14 خولەک خوێندنەوە · 12 stepsهەنگاو

What you will learnچی فێر دەبیت

  • Define military AI and name at least six kinds of job it does, from office work to weapons.
  • Explain that AI learns patterns from data and does one narrow job; it does not think like a human.
  • Give the four reasons armed forces want AI and the limit of each.
  • Explain dual use with an example.
  • Describe the course map (sense, understand, decide, act, protect) and why people stay in control.
  • پێناسە بکەیت و لانیکەم شەش جۆر کار ناو ببەیت کە دەیکات، لە کاری ئیدارییەوە تا چەک.
  • ڕوونی بکەیتەوە کە لە وێنەک فێر دەبێت و یەک کاری تەسک دەکات؛ وەک مرۆڤ بیر ناکاتەوە.
  • چوار هۆکارەکەی ویستنی لەلایەن هێزە چەکدارەکانەوە و سنووری هەریەکەیان بڵێیت.
  • بە نموونەیەک ڕوون بکەیتەوە.
  • نەخشەی کۆرسەکە (هەستکردن، تێگەیشتن، بڕیاردان، کارکردن، پاراستن) وەسف بکەیت و بۆچی مرۆڤ دەسەڵاتدار دەمێنێتەوە.

Military AI: a first definitionژیریی دەستکردی سەربازی: پێناسەیەکی سەرەتایی

Military AI is the use of artificial intelligence to help armed forces in their work. A short definition names four kinds of task: surveillance (watching an area over time), planning, fighting and protection. That is a fair start, but the real list is much longer, and most of it has nothing to do with weapons.

بەکارهێنانی بۆ یارمەتیدانی هێزە چەکدارەکان لە کارەکانیاندا. پێناسەیەکی کورت چوار جۆر کار ناو دەبات: (سەیرکردنی ناوچەیەک بۆ ماوەیەکی درێژ)، پلاندانان، شەڕ و پاراستن. ئەمە سەرەتایەکی باشە، بەڵام لیستە ڕاستەقینەکە زۆر درێژترە و زۆربەشی هیچ پەیوەندییەکی بە چەکەوە نییە.

Definitionپێناسە

The use of by armed forces for many tasks, from office work, and maintenance to training, cyber defence, and decision support. Only a small part of it is weapons, and people remain responsible for how it is used.

بەکارهێنانی لەلایەن هێزە چەکدارەکانەوە بۆ زۆر کار، لە کاری ئیداری، و تا ڕاهێنان، بەرگریی سایبەری، و پشتگیریی بڕیاردان. تەنیا بەشێکی بچووکی چەکە و مرۆڤەکان بەرپرسیاری چۆنیەتیی بەکارهێنانی دەمێننەوە.

A long bar of nine parts. From left to right: office work (pay, recruiting), logistics, maintenance, training, cyber defence, sensing (ISR), decision support, and vehicles and robots, each with a small picture. A bracket over these eight parts says: most military AI is not a weapon. A thin orange part at the far right is labelled: weapons, a small part. An arrow underneath runs from 'far from the use of force' to 'close to the use of force'. A note says the widths only illustrate.شریتێکی درێژ لە نۆ بەش. لە چەپەوە بۆ ڕاست: کاری ئیداری (مووچە، دامەزراندن)، لۆجستیک، چاککاری، ڕاهێنان، بەرگریی سایبەری، کۆکردنەوەی زانیاری (هەواڵگری، چاودێری و دیدەوانی)، پشتگیریی بڕیاردان و ئۆتۆمبێل و ڕۆبۆتەکان، هەریەکە بە وێنەیەکی بچووکەوە. کەوانەیەک لەسەر ئەم هەشت بەشە دەڵێت: زۆربەی ژیریی دەستکردی سەربازی چەک نییە. بەشێکی باریکی پرتەقاڵی لە کۆتایی لای ڕاست نووسراوە: چەک، بەشێکی بچووک. تیرێک لە خوارەوە لە «دوور لە بەکارهێنانی هێز»ەوە دەچێت بۆ «نزیک لە بەکارهێنانی هێز». تێبینییەک دەڵێت پانیی بەشەکان تەنیا بۆ ڕوونکردنەوەیە.
From payroll to weapons: most military AI does ordinary work.لە مووچەوە تا چەک: زۆربەی کاری ئاسایی دەکات.CêIT

Eight kinds of jobهەشت جۆر کار

Here are the eight kinds of job on the bar, each with an everyday example and the lesson of this course that teaches it. UNIDIR groups most of them under four big military tasks: command and control (deciding), information management (collecting and sharing information), logistics, and training.

ئەمانە هەشت جۆرە کارەکەی سەر شریتەکەن، هەریەکە لەگەڵ نموونەیەکی ڕۆژانە و ئەو وانەیەی ئەم کۆرسە کە فێری دەکات. UNIDIR زۆربەیان لە ژێر چوار ئەرکی گەورەی سەربازیدا کۆ دەکاتەوە: فەرماندەیی و کۆنترۆڵ (بڕیاردان)، بەڕێوەبردنی زانیاری (کۆکردنەوە و هاوبەشکردنی زانیاری)، و ڕاهێنان.

Military AI by kind of job بەپێی جۆری کار
JobکارExampleنموونەTaught inلە کوێ فێر دەکرێت
Office workکاری ئیداریSoftware checks salary payments or sorts job applications. پارەدانی مووچە دەپشکنێت یان داواکاریی دامەزراندن ڕیز دەکات.this lessonئەم وانەیە
LogisticsPlanning how fuel, food and spare parts reach the right place on time.پلاندانان بۆ ئەوەی سووتەمەنی، خواردن و پارچەی یەدەگ لە کاتی خۆیدا بگەنە شوێنی گونجاو.Logistics, maintenance and trainingلۆجستیک، چاککاری و ڕاهێنان
MaintenancePredicting which engine part will fail soon, so it is replaced in time.پێشبینیکردنی ئەوەی کام پارچەی بزوێنەر بەم زووانە تێکدەچێت، بۆ ئەوەی لە کاتی خۆیدا بگۆڕدرێت.Logistics, maintenance and trainingلۆجستیک، چاککاری و ڕاهێنان
TrainingڕاهێنانA simulator that makes an exercise harder or easier for each trainee. کە ڕاهێنانێک بۆ هەر ڕاهێنراوێک قورستر یان ئاسانتر دەکات.Logistics, maintenance and trainingلۆجستیک، چاککاری و ڕاهێنان
Cyber defenceبەرگریی سایبەریSpotting unusual activity on a military computer network.دۆزینەوەی چالاکیی نائاسایی لە سەربازیدا.AI in cyber defenceژیریی دەستکرد لە بەرگریی سایبەریدا
Sensing (ISR)کۆکردنەوەی زانیاری ()Sorting thousands of drone and satellite pictures.ڕیزکردنی هەزاران وێنەی و مانگی دەستکرد.Surveillance and reconnaissanceچاودێری و دیدەوانی
Decision supportپشتگیریی بڕیاردانShowing a commander options, with reasons and how certain each one is.پیشاندانی چەند هەڵبژاردەیەک بە فەرماندەیەک، لەگەڵ هۆکارەکان و ڕادەی دڵنیایی هەریەکەیان.Decisions, human control and responsibilityبڕیار، کۆنترۆڵی مرۆڤ و بەرپرسیارێتی
Vehicles and robotsئۆتۆمبێل و A robot that a person drives from a safe distance to examine a suspicious object. کە کەسێک لە دوورییەکی سەلامەتەوە لێی دەخوڕێت بۆ پشکنینی تەنێکی گوماناوی.Autonomous drones, vehicles and robotsدرۆن، ئۆتۆمبێل و ڕۆبۆتی سەربەخۆ
Weapon systemsسیستەمی چەکWeapons with AI functions: a small part, but the most debated, with strict legal limits.چەکی خاوەن ئەرکی : بەشێکی بچووکە، بەڵام زۆرترین گفتوگۆی لەسەرە و سنووری یاسایی توندی هەیە.War has rulesشەڕیش یاسای هەیە
Four boxes for UNIDIR's four big military tasks: command and control (deciding, decision support), information management (sensing and sharing information), logistics (moving supplies, maintenance) and training (simulators, exercises). Below: also office work such as pay and recruiting, and weapons as a very small portion; there is no agreed international list.چوار سندووق بۆ چوار ئەرکە گەورە سەربازییەکەی UNIDIR: فەرماندەیی و کۆنترۆڵ (بڕیاردان، پشتگیریی بڕیاردان)، بەڕێوەبردنی زانیاری (کۆکردنەوە و هاوبەشکردنی زانیاری)، لۆجستیک (گواستنەوەی کەلوپەل، چاککاری) و ڕاهێنان (هاوشێوەسازەکان، مەشق). لە خوارەوە: هەروەها کاری ئیداری وەک مووچە و دامەزراندن و چەک وەک بەشێکی زۆر بچووک؛ لیستێکی نێودەوڵەتیی ڕێککەوتوو نییە.
One useful way to group military AI.ڕێگایەکی بەسوود بۆ پۆلێنکردنی .CêIT

What counts as AI?چی بە ژیریی دەستکرد دادەنرێت؟

There is no single agreed definition of AI; What is AI? shows three classic ones. The 2023 Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy, which more than 50 countries had endorsed by 2025, uses a working definition: AI is the ability of machines to do tasks that would otherwise need human intelligence. It names five such abilities:

یەک پێناسەی ڕێککەوتوو بۆ نییە؛ ژیریی دەستکرد چییە؟ سێ پێناسەی بەناوبانگ پیشان دەدات. ڕاگەیەندراوی سیاسیی ساڵی ⁦2023⁩ لەبارەی بەکارهێنانی سەربازیی بەرپرسیارانەی و ، کە تا ساڵی ⁦2025⁩ زیاتر لە ⁦50⁩ وڵات پشتگیرییان لێ کردبوو، پێناسەیەکی کاری بەکار دەهێنێت: توانای ئامێرە بۆ ئەنجامدانی ئەو کارانەی کە ئەگەر ئامێر نەبوایە پێویستیان بە ژیریی مرۆڤ دەبوو. پێنج توانای لەم جۆرە ناو دەبات:

Five numbered cards. 1 Recognise patterns: spots a truck in a satellite picture. 2 Learn from experience: gets better after seeing more labelled pictures. 3 Draw conclusions: 'The bridge is flooded, so the trucks must wait.' 4 Make predictions: 'This engine will probably fail within 50 hours.' 5 Give recommendations: 'Take the north road: longer, but dry.' A note says this is a working definition from the 2023 Political Declaration and that there is no single agreed definition of AI.پێنج کارتی ژمارەدار. 1 ناسینەوەی وێنەک: بارهەڵگرێک لە وێنەیەکی مانگی دەستکرددا دەدۆزێتەوە. 2 فێربوون لە ئەزموون: دوای بینینی وێنەی نیشانەکراوی زیاتر باشتر دەبێت. 3 دەرەنجام وەرگرتن: «ئاو پردەکەی گرتووە، بۆیە بارهەڵگرەکان دەبێت چاوەڕێ بکەن.» 4 پێشبینی کردن: «ئەم بزوێنەرە لەوانەیە لە ماوەی 50 کاتژمێردا تێکبچێت.» 5 پێشنیار کردن: «ڕێگای باکوور بگرە: درێژترە، بەڵام وشکە.» تێبینییەک دەڵێت ئەمە پێناسەیەکی کارییە لە ڕاگەیەندراوی سیاسیی 2023 و یەک پێناسەی ڕێککەوتووی ژیریی دەستکرد نییە.
Five abilities, each with an everyday military example.پێنج توانا، هەریەکە لەگەڵ نموونەیەکی ڕۆژانەی سەربازی.CêIT

AI learns patterns; it does not think like usژیریی دەستکرد وێنەک فێر دەبێت، وەک ئێمە بیر ناکاتەوە

AI is often described as something that “thinks and learns like a human, in limited ways”. The second half matters most. Today's AI does not think like a person. It learns patterns from many examples (this is machine learning) and then applies them. Each system does one narrow job: this is narrow AI, explained in What is AI?.

زۆرجار وا وەسف دەکرێت کە «وەک مرۆڤ بیر دەکاتەوە و فێر دەبێت، بەڵام بە شێوەیەکی سنووردار». نیوەی دووەمی ئەم ڕستەیە گرنگترە. ئەمڕۆ وەک مرۆڤ بیر ناکاتەوە. لە ژمارەیەکی زۆر نموونەوە وێنەک فێر دەبێت (ئەمە )، پاشان بەکاریان دەهێنێت. هەر سیستەمێک تەنیا یەک کاری تەسک دەکات: ئەمە کە لە وانەی ژیریی دەستکرد چییە؟دا ڕوون کراوەتەوە.

So what does it do? It helps machines and software make decisions inside one task: a robot picks a path around a rock, or a program ranks which pictures to look at first. People set the goal and check the result. And it processes huge amounts of data quickly: far more pictures, signals or records than a team could read. How this learning works is in Machine learning basics.

کەواتە چی دەکات؟ یارمەتی ئامێر و دەدات لە ناو یەک کاردا بڕیار بدەن: ڕێگایەک بە دەوری بەردێکدا هەڵدەبژێرێت، یان ڕیز دەکات کە یەکەم جار سەیری کام وێنانە بکرێت. مرۆڤ ئامانجەکە دیاری دەکات و ئەنجامەکە دەپشکنێت. هەروەها بڕێکی زۆر بە خێرایی شی دەکاتەوە: وێنە، یان تۆماری زۆر زیاتر لەوەی تیمێک بتوانێت بیخوێنێتەوە. چۆنیەتیی ئەم فێربوونە لە وانەی بنەماکانی فێربوونی ئامێردا ڕوون کراوەتەوە.

Left: a person who drives, cooks and talks, learns a new job, notices what is strange, explains why and knows what they do not know. Right: a narrow AI model that knows only three labels (truck, car, bus) is shown a picture of a camel and answers 'truck, 0.55'. Text: it must choose one of its own labels, even for a camel it has never seen. Footer: AI learns patterns from examples; it does not think like a person.لای چەپ: مرۆڤێک کە لێدەخوڕێت، چێشت لێدەنێت و قسە دەکات، کارێکی نوێ فێر دەبێت، هەست بە شتی سەیر دەکات، ڕوون دەکاتەوە بۆچی و دەزانێت چی نازانێت. لای ڕاست: مۆدێلێکی ژیریی دەستکردی تایبەت کە تەنیا سێ نیشانە دەزانێت (بارهەڵگر، ئۆتۆمبێل، پاس) وێنەی وشترێکی پیشان دەدرێت و وەڵام دەداتەوە «بارهەڵگر، 0.55». دەقەکە: دەبێت یەکێک لە نیشانەکانی خۆی هەڵبژێرێت، تەنانەت بۆ وشترێک کە هەرگیز نەیبینیوە. لە خوارەوە: ژیریی دەستکرد لە نموونەکانەوە ناسینەوەی وێنەک فێر دەبێت و وەک مرۆڤ بیر ناکاتەوە.
A narrow model must answer with one of its own labels, even for something it has never seen. تەسک دەبێت بە یەکێک لە خۆی وەڵام بداتەوە، تەنانەت بۆ شتێک کە هەرگیز نەیبینیوە.CêIT

Why armed forces want AIبۆچی هێزە چەکدارەکان ژیریی دەستکردیان دەوێت

Four reasons are given again and again:

چوار هۆکار دووپاتە و سێبارە باس دەکرێن:

  1. Faster decisions: AI does not get tired and can work very fast.
  2. More accuracy: it spots patterns and details that people may miss.
  3. Saves lives: robots can go where it is dangerous for soldiers.
  4. Handles big data: it can scan thousands of images or signals quickly.
  1. بڕیاری خێراتر: ماندوو نابێت و دەتوانێت زۆر خێرا کار بکات.
  2. وردیی زیاتر: وێنەک و وردەکاری دەبینێت کە لەوانەیە مرۆڤ نەیانبینێت.
  3. ژیان دەپارێزێت: دەتوانن بچنە ئەو شوێنانەی بۆ سەرباز مەترسیدارن.
  4. بەڕێوە دەبات: دەتوانێت هەزاران وێنە یان بە خێرایی بپشکنێت.

All four are real. But each has a “but”: a limit that decides whether the benefit is safe. The next two screens explain them.

هەر چوارێکیان ڕاستن. بەڵام هەریەکەیان «بەڵام»ێکی هەیە: سنوورێک کە دیاری دەکات ئایا ئەو سوودە سەلامەتە یان نا. دوو پەڕەی داهاتوو ڕوونیان دەکەنەوە.

Four cards. Faster decisions: never gets tired, answers in seconds; but less time for people to check, and a mistake can escalate fast. More accuracy: spots patterns and details people may miss; but only in the conditions it was trained and tested for. Saves lives: robots go where it is dangerous for soldiers; but most are still driven by a person from a distance, and they can fail. Handles big data: scans thousands of images or signals quickly; but what it finds is a clue, and people must still judge it.چوار کارت. بڕیاری خێراتر: ماندوو نابێت و لە چەند چرکەیەکدا وەڵام دەداتەوە؛ بەڵام کاتی کەمتر بۆ پشکنینی مرۆڤ دەمێنێتەوە و هەڵەیەک دەتوانێت بە خێرایی ململانێ گەورە بکات. وردیی زیاتر: وێنەک و وردەکاری دەبینێت کە لەوانەیە مرۆڤ نەیبینێت؛ بەڵام تەنیا لەو بارودۆخانەدا کە تێیدا ڕاهێنراوە و تاقیکراوەتەوە. ژیان دەپارێزێت: ڕۆبۆتەکان دەچنە شوێنی مەترسیدار؛ بەڵام زۆربەیان هێشتا مرۆڤ لە دوورەوە لێیان دەخوڕێت و ئەوانیش تێکدەچن. دراوەی گەورە بەڕێوە دەبات: هەزاران وێنە یان سیگناڵ بە خێرایی دەپشکنێت؛ بەڵام ئەوەی دەیدۆزێتەوە سەرەداوە و مرۆڤ هێشتا دەبێت هەڵیسەنگێنێت.
Each benefit comes with a limit.هەر سوودێک سنوورێکی لەگەڵدایە.CêIT

Why armed forces want AIبۆچی هێزە چەکدارەکان ژیریی دەستکردیان دەوێت

Faster and more accurate, but…خێراتر و وردتر، بەڵام…

Speed. AI never gets tired, and it answers in seconds. That is a real gain when it sorts data. But a fast answer leaves less time for a person to check it. And when machines on two sides react to each other at machine speed, a small error can grow into a crisis before anyone can stop it. The UN General Assembly lists “miscalculations” and the “escalation of conflicts” among its worries about autonomous weapons (resolution 79/62, December 2024).

خێرایی. هەرگیز ماندوو نابێت و لە چەند چرکەیەکدا وەڵام دەداتەوە. ئەمە کاتێک ڕیز دەکات سوودێکی ڕاستەقینەیە. بەڵام وەڵامێکی خێرا کاتی کەمتر بۆ پشکنینی مرۆڤ دەهێڵێتەوە. کاتێکیش ئامێرەکانی دوو لایەن بە خێرایی ئامێر کاردانەوە بەرانبەر بە یەکتر دەنوێنن، هەڵەیەکی بچووک دەتوانێت ببێتە قەیرانێک پێش ئەوەی کەس بتوانێت ڕایبگرێت. کۆبوونەوەی گشتیی نەتەوە یەکگرتووەکان «هەڵەی خەمڵاندن» و «پەرەسەندنی ململانێکان» لە نێو نیگەرانییەکانیدا لەبارەی چەکی ناو دەبات (بڕیاری ⁦79/62⁩، کانوونی یەکەمی ⁦2024⁩).

Accuracy. AI can beat people on the exact task it was trained and tested for. In new conditions (fog, snow, another country, camouflage) it can be much worse, and nobody may notice. So accuracy is not a promise but a measurement: how many false alarms and how many misses it makes in realistic tests. When AI gets it wrong shows how this is measured.

وردی. دەتوانێت لەو کارە دیاریکراوەدا کە بۆی ڕاهێنراوە و تاقیکراوەتەوە لە مرۆڤ باشتر بێت. لە بارودۆخی نوێدا (تەم، بەفر، وڵاتێکی تر، ) دەکرێت زۆر خراپتر بێت و لەوانەیە کەس هەستی پێ نەکات. بۆیە وردی بەڵێن نییە، پێوانەیە: لە تاقیکردنەوەی نزیک لە ڕاستیدا چەند ئاگادارکردنەوەی هەڵە و چەند لەدەستدان دەکات. وانەی کاتێک ژیریی دەستکرد هەڵە دەکات پیشان دەدات چۆن دەپێورێت.

Two panels. Left, 'Helps: sorting a big pile of pictures': for 10,000 pictures (an example), people alone need days; with AI sorting first, it takes minutes, then people check the top few. If a picture waits a little, nobody is harmed. Right, 'Hurts: fast systems react to each other': two machines send arrows back and forth; a small error can grow in seconds, while a person needs minutes to check and stop it. Bottom: where harm is possible, keep a human check.دوو بەش. لای چەپ، «یارمەتی دەدات: ڕیزکردنی کۆمەڵێکی زۆر وێنە»: بۆ 10,000 وێنە (نموونەیەک)، تەنیا مرۆڤ چەند ڕۆژێکی پێویستە؛ ئەگەر ژیریی دەستکرد سەرەتا ڕیزیان بکات، چەند خولەکێک دەخایەنێت و پاشان مرۆڤ سەرەوەکان دەپشکنێت. ئەگەر وێنەیەک کەمێک چاوەڕێ بکات، زیان بە کەس ناگات. لای ڕاست، «زیان دەگەیەنێت: سیستەمە خێراکان کاردانەوە بەرانبەر یەکتر دەنوێنن»: دوو ئامێر تیر بۆ یەکتر دەنێرن؛ هەڵەیەکی بچووک لە چەند چرکەیەکدا گەورە دەبێت، لە کاتێکدا مرۆڤ چەند خولەکێکی پێویستە بۆ پشکنین و ڕاگرتنی. لە خوارەوە: لەو شوێنانەی زیان دەکرێت، پشکنینی مرۆڤ بهێڵەوە.
Speed is a gain for sorting, and a risk for decisions nobody has time to check.خێرایی بۆ ڕیزکردن سوودە و بۆ ئەو بڕیارانەی کەس کاتی پشکنینیانی نییە مەترسییە.CêIT

Why armed forces want AIبۆچی هێزە چەکدارەکان ژیریی دەستکردیان دەوێت

Safer and bigger, but…سەلامەتتر و گەورەتر، بەڵام…

Saving lives. Robots can go where it is dangerous for people. A bomb-disposal robot examines a suspicious object while its operator stays at a safe distance. After an earthquake, snake robots can search inside collapsed buildings (AI that protects). Most of these robots are driven by a person from far away; AI helps them see and move smoothly (Autonomous drones, vehicles and robots). They can still break down or misread a scene, so people plan for failures.

پاراستنی ژیان. دەتوانن بچنە ئەو شوێنانەی بۆ مرۆڤ مەترسیدارن. تەنێکی گوماناوی دەپشکنێت، لە کاتێکدا ئۆپەراتۆرەکەی لە دوورییەکی سەلامەتدا دەمێنێتەوە. دوای بوومەلەرزە، دەتوانن لە ناو باڵەخانە ڕووخاوەکاندا بگەڕێن (ژیریی دەستکرد بۆ پاراستن). زۆربەی ئەم مرۆڤ لە دوورەوە لێیان دەخوڕێت؛ یارمەتییان دەدات ببینن و بە نەرمی بجووڵێن (درۆن، ئۆتۆمبێل و ڕۆبۆتی سەربەخۆ). هێشتا دەکرێت تێکبچن یان دیمەنێک بە هەڵە بخوێننەوە، بۆیە مرۆڤ پلان بۆ تێکچوونیان دادەنێت.

Big data. AI can scan thousands of images or signals in minutes and put the few that matter on top of the pile. But what it flags is a clue, not a fact: a trained person must still judge it. How this sorting (triage) works is taught in Surveillance and reconnaissance.

. دەتوانێت لە چەند خولەکێکدا هەزاران وێنە یان بپشکنێت و ئەو چەند دانەیەی گرنگن بخاتە پێشەوە. بەڵام ئەوەی نیشانی دەکات سەرەداوە، نەک ڕاستییەکی سەلمێنراو: کەسێکی ڕاهێنراو هێشتا دەبێت هەڵیسەنگێنێت. چۆنیەتیی ئەم ڕیزکردنە () لە وانەی چاودێری و دیدەوانی فێر دەکرێت.

Dual use: one technology, two purposesدوو بەکارهێنان: یەک تەکنەلۆژیا، دوو مەبەست

Most AI is built for civilian life, and the same tools work for armed forces. This is called dual use. A computer-vision model that finds vehicles in satellite pictures can be retrained to find flooded roads after a storm. In fact, UNOSAT, the UN's satellite centre, uses AI to map floods from radar satellite pictures and sends the maps to relief teams (Surveillance and reconnaissance).

زۆربەی بۆ ژیانی مەدەنی دروست دەکرێت و هەمان ئامرازەکان بۆ هێزە چەکدارەکانیش کار دەکەن. پێی دەوترێت . کە ئۆتۆمبێل لە دەدۆزێتەوە، دەکرێت دووپاتە ڕابهێنرێت بۆ دۆزینەوەی ڕێگا ژێرئاوکەوتووەکان دوای زریانێک. لە ڕاستیدا UNOSAT، ناوەندی مانگی دەستکردی نەتەوە یەکگرتووەکان، بەکار دەهێنێت بۆ نەخشەکێشانی لافاو لە وێنەکانی مانگی دەستکردەوە و نەخشەکان بۆ تیمەکانی فریاکەوتن دەنێرێت (چاودێری و دیدەوانی).

Dual use also works the other way: a model built for the military can be reused for rescue work. Drones are the best-known case: armies built the first ones, as training targets and for reconnaissance, and today drones map fields and search for missing people (A short history of drones). Because the same AI can help or harm, the people who build, sell and use it must ask who will use it, for what, and with which checks.

بە پێچەوانەشەوە ڕوو دەدات: کە بۆ سوپا دروست کراوە، دەکرێت بۆ کاری فریاکەوتن بەکار بهێنرێتەوە. ناسراوترین نموونەن: یەکەمینەکانیان سوپاکان وەک ئامانجی ڕاهێنان و بۆ دروستیان کردن و ئەمڕۆ نەخشەی کێڵگە دەکێشن و بە دوای کەسانی ونبوودا دەگەڕێن (مێژوویەکی کورتی درۆن). چونکە هەمان دەتوانێت یارمەتی بدات یان زیان بگەیەنێت، ئەوانەی دروستی دەکەن، دەیفرۆشن و بەکاری دەهێنن دەبێت بپرسن: کێ بەکاری دەهێنێت، بۆ چی و بە چ پشکنینێک؟

The same drawn satellite picture twice: a river, two roads, houses and small vehicles. Left, 'Flood relief: which roads are under water?': a dashed blue area marks the flooded land. Right, 'Military use: where are the vehicles?': orange boxes mark each vehicle. In the middle: the same picture, the same kind of model. Bottom: makers and users must think about both uses, and check who will use it and how. The picture is a drawing, not a real satellite picture.هەمان وێنەی مانگی دەستکرد کێشراو دوو جار: ڕووبارێک، دوو ڕێگا، خانوو و ئۆتۆمبێلی بچووک. لای چەپ، «فریاکەوتنی لافاو: کام ڕێگا لە ژێر ئاودایە؟»: ناوچەیەکی شینی هێڵ پچڕپچڕ زەوییە ژێرئاوکەوتووەکە نیشان دەدات. لای ڕاست، «بەکارهێنانی سەربازی: ئۆتۆمبێلەکان لە کوێن؟»: چوارگۆشەی پرتەقاڵی هەر ئۆتۆمبێلێک نیشان دەدات. لە ناوەڕاستدا: هەمان وێنە، هەمان جۆری مۆدێل. لە خوارەوە: دروستکەر و بەکارهێنەران دەبێت بیر لە هەردوو بەکارهێنانەکە بکەنەوە و بزانن کێ و چۆن بەکاری دەهێنێت. وێنەکە کێشراوە، نەک وێنەی مانگی دەستکرد ڕاستەقینە.
One model, two purposes.یەک ، دوو مەبەست.CêIT

The map of this courseنەخشەی ئەم کۆرسە

This course follows the path information takes through a system: it is sensed, then understood, then someone decides, then a machine may act. And the same technology can protect people. Each step has its lessons:

ئەم کۆرسە ئەو ڕێگایە دەگرێتەبەر کە زانیاری بە ناو سیستەمێکدا دەیبڕێت: سەرەتا هەستی پێ دەکرێت، پاشان لێی تێدەگەن، پاشان کەسێک بڕیار دەدات، پاشان لەوانەیە ئامێرێک کار بکات. هەمان تەکنەلۆژیاش دەتوانێت خەڵک بپارێزێت. هەر هەنگاوێک وانەکانی خۆی هەیە:

Animation: five boxes in a loop, Sense, Understand, Decide, Act and Protect, light up one after another while the middle shows each step's lessons. In the last frame all five are lit and the middle says: people stay in control of every step.
Animation: five boxes in a loop, Sense, Understand, Decide, Act and Protect, light up one after another while the middle shows each step's lessons. In the last frame all five are lit and the middle says: people stay in control of every step.
Sense, understand, decide, act, protect: people stay in control.هەستکردن، تێگەیشتن، بڕیاردان، کارکردن، پاراستن: مرۆڤ دەسەڵاتدار دەمێنێتەوە.CêIT

Common mistakesهەڵە باوەکان

MistakeهەڵەBetterباشتر
“Military AI means killer robots.”« واتە کوشندە.»Most military AI does office work, logistics, maintenance, training, cyber defence and sensing. Weapons are a small part.زۆربەی کاری ئیداری، ، ، ڕاهێنان، بەرگریی سایبەری و کۆکردنەوەی زانیاری دەکات. چەک بەشێکی بچووکە.
“AI thinks like a human.”« وەک مرۆڤ بیر دەکاتەوە.»AI learns patterns from examples and does one narrow job. It does not know what it does not know. لە نموونەکانەوە وێنەک فێر دەبێت و یەک کاری تەسک دەکات. نازانێت چی نازانێت.
“AI is always more accurate than people.”« هەمیشە لە مرۆڤ وردترە.»Only on the task and in the conditions it was tested for. Accuracy is measured with false alarms and misses.تەنیا لەو کار و بارودۆخەدا کە تێیدا تاقیکراوەتەوە. وردی بە ئاگادارکردنەوەی هەڵە و لەدەستدانەکان دەپێورێت.
“Faster is always better.”«خێراتر هەمیشە باشترە.»Fast sorting helps. Decisions faster than people can check them are a risk.ڕیزکردنی خێرا یارمەتی دەدات. ئەو بڕیارانەی لەوەی مرۆڤ بتوانێت بیانپشکنێت خێراترن مەترسین.
“If the AI makes a mistake, the AI is responsible.”«ئەگەر هەڵە بکات، بەرپرسیارە.»Responsibility cannot be passed to a machine: people and states stay responsible (Decisions, human control and responsibility).بەرپرسیارێتی ناگوازرێتەوە بۆ ئامێر: مرۆڤ و دەوڵەتەکان بەرپرسیار دەمێننەوە (بڕیار، کۆنترۆڵی مرۆڤ و بەرپرسیارێتی).

Exercisesڕاهێنانەکان

ExerciseڕاهێنانSort the examplesنموونەکان ڕیز بکە

Put each example into one part of the bar (office work, logistics, maintenance, training, cyber defence, sensing, decision support, vehicles and robots): (a) a program that warns that a generator will overheat next week; (b) a chatbot that answers soldiers' questions about their salary; (c) software that ranks three possible routes for a medical convoy; (d) a tool that sorts 5,000 drone pictures after a flood; (e) a simulator that makes an exercise harder when the trainee does well; (f) a system that spots strange logins on a base network.

هەر نموونەیەک بخەرە یەکێک لە بەشەکانی شریتەکە (کاری ئیداری، ، ، ڕاهێنان، بەرگریی سایبەری، کۆکردنەوەی زانیاری، پشتگیریی بڕیاردان، ئۆتۆمبێل و ): ⁦(a)⁩ کە ئاگادار دەکاتەوە مۆلیدەیەک هەفتەی داهاتوو زۆر گەرم دەبێت؛ ⁦(b)⁩ کە وەڵامی پرسیاری سەربازەکان لەبارەی مووچەکەیان دەداتەوە؛ ⁦(c)⁩ کە سێ ڕێگای ئەگەری بۆ کاروانێکی پزیشکی ڕیز دەکات؛ ⁦(d)⁩ ئامرازێک کە دوای لافاو ⁦5,000⁩ وێنەی ڕیز دەکات؛ ⁦(e)⁩ کە کاتێک ڕاهێنراوەکە باش دەکات، ڕاهێنانەکە قورستر دەکات؛ ⁦(f)⁩ سیستەمێک کە سەیر لە بنکەیەکدا دەدۆزێتەوە.

Hintئاماژە

Ask: what is the tool for, and who uses its answer?

بپرسە: ئامرازەکە بۆ چییە و کێ وەڵامەکەی بەکار دەهێنێت؟

Show answerوەڵامەکە پیشان بدە
ExampleنموونەPart of the barبەشی شریتەکە
(a)Maintenance (predicting a fault) (پێشبینیکردنی تێکچوون)
(b)Office workکاری ئیداری
(c)Decision support (it ranks options for a person); route planning for supplies is also logisticsپشتگیریی بڕیاردان (هەڵبژاردەکان بۆ کەسێک ڕیز دەکات)؛ پلاندانانی ڕێگا بۆ کەلوپەل
(d)Sensing (here used for flood relief)کۆکردنەوەی زانیاری (لێرەدا بۆ فریاکەوتنی لافاو)
(e)Trainingڕاهێنان
(f)Cyber defenceبەرگریی سایبەری
ExerciseڕاهێنانFind the “but”«بەڵام»ەکە بدۆزەرەوە

A news report says: “This AI is 98% accurate, never gets tired and decides in one second, so it is safer than people.” Write three questions you would ask before believing it.

هەواڵێک دەڵێت: «ئەم ⁦98%⁩ وردە، هەرگیز ماندوو نابێت و لە یەک چرکەدا بڕیار دەدات، بۆیە لە مرۆڤ سەلامەتترە.» سێ پرسیار بنووسە کە پێش باوەڕکردن پێی دەیپرسیت.

Show answerوەڵامەکە پیشان بدە
  • 98% on which test data, and in which conditions (night, bad weather, another country)?
  • How many false alarms and how many misses are inside that 98%, and which one is worse here?
  • Who checks a decision made in one second, how much time do they have, and who is responsible when it is wrong?
  • ⁦98%⁩ لەسەر چ و لە چ بارودۆخێکدا (شەو، کەشوهەوای خراپ، وڵاتێکی تر)؟
  • چەند ئاگادارکردنەوەی هەڵە و چەند لەدەستدان لە ناو ئەو ⁦98%⁩ـەدایە و لێرەدا کامیان خراپترە؟
  • کێ بڕیارێک دەپشکنێت کە لە یەک چرکەدا دراوە، چەند کاتی هەیە و کاتێک هەڵە بێت کێ بەرپرسیارە؟
ExerciseڕاهێنانDual useدوو بەکارهێنان

A company sells a drone with a thermal camera and software that finds people at night. Give one peaceful use, one military use, and one question the company should ask before selling it.

کۆمپانیایەک بە و دەفرۆشێت کە شەوانە خەڵک دەدۆزێتەوە. یەک بەکارهێنانی ئاشتیانە، یەک بەکارهێنانی سەربازی و یەک پرسیار بنووسە کە کۆمپانیاکە پێش فرۆشتن دەبێت بیپرسێت.

Show answerوەڵامەکە پیشان بدە

Peaceful: finding lost hikers or survivors after an earthquake. Military: watching a border at night. Question: who is the buyer, what will it be used for, and does that use follow the law and protect civilians?

ئاشتیانە: دۆزینەوەی شاخەوانی ونبوو یان ڕزگاربووان دوای بوومەلەرزە. سەربازی: سنوور بە شەو. پرسیار: کڕیارەکە کێیە، بۆ چی بەکاری دەهێنێت و ئایا ئەو بەکارهێنانە پەیڕەوی یاسا دەکات و هاووڵاتیانی مەدەنی دەپارێزێت؟

Summaryپوختە

  • Military AI is the use of AI by armed forces: from pay, logistics and maintenance to training, cyber defence, sensing, decision support, robots and, as a small part, weapons.
  • There is no single agreed definition of AI. A common working one: machines doing tasks that would otherwise need human intelligence (recognising patterns, learning, concluding, predicting, recommending).
  • AI learns patterns from data and does one narrow job; it does not think like a person.
  • Four reasons (speed, accuracy, saving lives, big data) each have a limit: time to check, tested conditions only, failures, human judgement.
  • Dual use: the same AI can protect or harm, so makers and users must think about both.
  • At every step of the course map, people stay in control and responsible.
  • بەکارهێنانی لەلایەن هێزە چەکدارەکانەوە: لە مووچە، و تا ڕاهێنان، بەرگریی سایبەری، کۆکردنەوەی زانیاری، پشتگیریی بڕیاردان، و وەک بەشێکی بچووک، چەک.
  • یەک پێناسەی ڕێککەوتووی نییە. پێناسەیەکی کاریی باو: ئامێر ئەو کارانە دەکات کە ئەگەر ئامێر نەبوایە پێویستیان بە ژیریی مرۆڤ دەبوو (، فێربوون، دەرەنجام وەرگرتن، پێشبینی، پێشنیار).
  • لە وێنەک فێر دەبێت و یەک کاری تەسک دەکات؛ وەک مرۆڤ بیر ناکاتەوە.
  • چوار هۆکارەکە (خێرایی، وردی، پاراستنی ژیان، ) هەریەکەیان سنوورێکی هەیە: کاتی پشکنین، تەنیا بارودۆخی تاقیکراوە، تێکچوون، هەڵسەنگاندنی مرۆڤ.
  • : هەمان دەتوانێت بپارێزێت یان زیان بگەیەنێت، بۆیە دروستکەر و بەکارهێنەران دەبێت بیر لە هەردووکیان بکەنەوە.
  • لە هەموو هەنگاوێکی نەخشەی کۆرسەکەدا مرۆڤ دەسەڵاتدار و بەرپرسیار دەمێنێتەوە.
Review questionsپرسیارەکانی پێداچوونەوە
Is most military AI part of weapons?ئایا زۆربەی ژیریی دەستکردی سەربازی بەشێکە لە چەک؟

No. Most of it does office work, logistics, maintenance, training, cyber defence and sensing. UNIDIR calls autonomy in weapons a very small portion of the possible uses.

نەخێر. زۆربەی کاری ئیداری، ، ، ڕاهێنان، بەرگریی سایبەری و کۆکردنەوەی زانیاری دەکات. UNIDIR چەک بە بەشێکی زۆر بچووکی بەکارهێنانە ئەگەرییەکان دادەنێت.

Why is “AI thinks like a human” misleading?بۆچی «ژیریی دەستکرد وەک مرۆڤ بیر دەکاتەوە» گومڕاکەرە؟

AI learns statistical patterns from examples and does one narrow task. It cannot explain itself like a person and does not know when a picture is outside what it learned.

لە نموونەکانەوە وێنەکی ئاماری فێر دەبێت و یەک کاری تەسک دەکات. ناتوانێت وەک مرۆڤ خۆی ڕوون بکاتەوە و نازانێت کەی وێنەیەک لە دەرەوەی ئەوەیە کە فێری بووە.

What is the limit of “faster decisions”?سنووری «بڕیاری خێراتر» چییە؟

Less time for people to check, and machines reacting to machines can turn a small error into a crisis.

کاتی کەمتر بۆ پشکنینی مرۆڤ و کاردانەوەی ئامێر بۆ ئامێر دەتوانێت هەڵەیەکی بچووک بکاتە قەیران.

Lesson complete!وانەکە تەواو بوو!

Well done. Review the key terms, then go on to the next lesson.دەستخۆش. سەیرێکی زاراوە سەرەکییەکان بکەرەوە، پاشان بڕۆ بۆ وانەی داهاتوو.

Key termsزاراوە سەرەکییەکان

Next lessonوانەی داهاتوو

How AI Sees: Sensors, Computer Vision and Detectionژیریی دەستکرد چۆن دەبینێت: هەستەوەرەکان، بینایی کۆمپیوتەر و دۆزینەوە

References and further readingسەرچاوەکان و خوێندنەوەی زیاتر

This lecture is a first edition. The Kurdish text and terminology are waiting for review by Kurdish IT teachers.ئەم وانەیە چاپی یەکەمە. دەقە کوردییەکە و زاراوەکانی چاوەڕێی پێداچوونەوەی مامۆستایانی کوردی بواری IT ن.

Appearanceڕووکار

Modeدۆخ
Styleشێواز
Coloursڕەنگەکان

Saved on this device only.تەنیا لەسەر ئەم ئامێرە پاشەکەوت دەکرێت.