How does artificial intelligence learn? - Briana Brownell

Translator: Dilnoza Nishanova Reviewer: Nazarbek Nazarov

Today, artificial intelligence helps doctors diagnose patients,

Bugun sun’iy intellekt shifokorlarga tashxis qo’yishda yordam bermoqda,

pilots fly commercial aircraft, and city planners predict traffic.

pilotlarga samolyot boshqarishda, muhandis -larga tirbandliklarni prognoz qilishda.

But no matter what these AIs are doing, the computer scientists who designed them

Lekin SI nima qilishidan qat’iy nazar, uni ishlab chiqqan dasturchilar

likely don’t know exactly how they’re doing it.

katta ehtimol bilan qanday qilayotganini bilmaydilar.

This is because artificial intelligence is often self-taught,

Buning sababi ko’p hollarda SI o’zi o’rgangan,

working off a simple set of instructions

oddiygina qo’llanmalar to’plamidan

to create a unique array of rules and strategies.

o’zgacha qoida va strategiyalar yaratadi.

So how exactly does a machine learn?

Xo’sh aynan qanday qilib mashina o’rganadi?

There are many different ways to build self-teaching programs.

O’zi o’rganadigan programmalar yaratishning ko’p yo’llari bor.

But they all rely on the three basic types of machine learning:

Lekin ularning barchasi uch asosiy mashina o’rganuvchi sistemaga asoslanadi:

unsupervised learning, supervised learning, and reinforcement learning.

nazoratsiz o’rganish, nazoratli o’rganish va mustahkam o’rganish.

To see these in action,

Bularni hayotda ko’rish uchun,

let’s imagine researchers are trying to pull information

tadqiqotchilarni minglab bemorlar tibbiy ma’lumotlari orasidan

from a set of medical data containing thousands of patient profiles.

informatsiya olishini tasavvur qilaylik.

First up, unsupervised learning.

Birinchi o’rinda, nazoratsiz o’rganish.

This approach would be ideal for analyzing all the profiles

Bu usul hamma profillarni umumiy o’xshashlik va

to find general similarities and useful patterns.

foydali yo’llarni topish uchun idealdir.

Maybe certain patients have similar disease presentations,

Balkim ba’zi bemorlarda o’xshash kasallik belgilar bor

or perhaps a treatment produces specific sets of side effects.

yoki ehtimol davolanishning ba’zi nuqsonlari bo’ladi.

This broad pattern-seeking approach can be used to identify similarities

Bu keng yo’l topuvchi usul inson aralashuvisiz

between patient profiles and find emerging patterns,

bemorlarning ma’lumotlaridan o’xshashlikni

all without human guidance.

topishda ishlatish mumkin.

But let's imagine doctors are looking for something more specific.

Lekin tasavvur qiling shifokorlar aniqroq narsani izlashmoqda.

These physicians want to create an algorithm

Bu shifokorlar ma’lum bir kasallikka

for diagnosing a particular condition.

tashxis qo’yish uchun algoritm yaratsishmoqchi.

They begin by collecting two sets of data—

Ular 2 to’plam ma’lumotlarni yig’ishdan boshlashadi--

medical images and test results from both healthy patients

ikkala sog’ bemorlar va kasallikka chalingan bemorlardan

and those diagnosed with the condition.

tibbiy rasmlar va test natijalarini olishadi.

Then, they input this data into a program

So’ng, bu ma’lumotlarni kasal bemorlarning

designed to identify features shared by the sick patients

xususiyatlarini aniqlash uchun dasturga kirg’iziladi

but not the healthy patients.

lekin sog’ bemorlarniki emas.

Based on how frequently it sees certain features,

Ma’lum xususiyatlarni qanchalik tez uchratishiga qarab,

the program will assign values to those features’ diagnostic significance,

dastur har bir xususiyatning muhimligiga qarab diagnostik qymat berib chiqadi,

generating an algorithm for diagnosing future patients.

va bemorlarga tashxis qo’yuvchi algoritm ishlab chiqaradi.

However, unlike unsupervised learning,

Ammo, boshqaruvsiz o’rganishdan farqli ravishda,

doctors and computer scientists have an active role in what happens next.

shifokor va dasturchilar keyingi jarayonda faol rol o’ynaydilar.

Doctors will make the final diagnosis

Shifokorlar oxirgi tashxisni qo’yadilar va

and check the accuracy of the algorithm’s prediction.

algoritm natijasining aniqligini tekshiradilar.

Then computer scientists can use the updated datasets

Dasturchilar yangi ma’lumotlarni dastur

to adjust the program’s parameters and improve its accuracy.

paramaterlarini to’g’irlash va aniqligini yaxshilash uchun ishlatishlari

This hands-on approach is called supervised learning.

Bu amaliy usul boshqaruvsiz o’rganish deyiladi.

Now, let’s say these doctors want to design another algorithm

Endi, ayataylik shifokorlar boshqa davolash rejasi

to recommend treatment plans.

algoritm tuzmoqchilar.

Since these plans will be implemented in stages,

Bu rejalar qadamma-qadam borar ekan va har bir

and they may change depending on each individual's response to treatments,

bemorning holatiga qarab o’zgarar ekan, shifokorlar

the doctors decide to use reinforcement learning.

mustahkamlovchi o’rganishni ishlatishga qaror qilganlar.

This program uses an iterative approach to gather feedback

Bu dastur takroriy usulni qaysi dori va davolash

about which medications, dosages and treatments are most effective.

eng optimali ekanligini aniqlashda feedbek to’playdi.

Then, it compares that data against each patient’s profile

Keyin, har bir bemor bilan optimal bo’lgan

to create their unique, optimal treatment plan.

davolash rejasini tuzish uchun solishtirib chiqadi.

As the treatments progress and the program receives more feedback,

Davolash davom etar va dastur feedback olar ekan,

it can constantly update the plan for each patient.

u doimiy tarzda har bir bemor rejasini yangilab boradi.

None of these three techniques are inherently smarter than any other.

Hech qaysi uch usul bir-biridan kelib chiqishi jihatidan aqlliroq emas.

While some require more or less human intervention,

Ba’zilari ko’proq yo kamroq inson aralashuvini qilsa-da,

they all have their own strengths and weaknesses

barchasini ma’lum ishga mos qiladigan

which makes them best suited for certain tasks.

kuchli va kuchsiz jihatlari bor.

However, by using them together,

Lekin, ularni birga ishlatib, tadqiqotchilar

researchers can build complex AI systems,

mukammal ma’lum dasturlar, bir-birini nazorat qiladigan va

where individual programs can supervise and teach each other.

o’rgatadigan SI sistemalar yarata oladilar.

For example, when our unsupervised learning program

Misol uchun, boshqaruvsiz o’rganadigan dastur

finds groups of patients that are similar,

o’xshash bemorlarni guruhini topsa,

it could send that data to a connected supervised learning program.

buni ulangan boshqaruvli o’rganish dasturiga yuboradi.

That program could then incorporate this information into its predictions.

U dastur keyin bu ma’lumotni bashoratga aylantira oladi

Or perhaps dozens of reinforcement learning programs

Yoki ehtimol ko’plab majburiy o’rganishning ma’lumotlari

might simulate potential patient outcomes

turli davolash rejalari haqidagi mulohazalarni

to collect feedback about different treatment plans.

ehtimoliy bemorning oqibalaridan to’play oladilar.

There are numerous ways to create these machine-learning systems,

O’zi o’rganadigan mashinalar yaratishning bir qancha yo’llari bor,

and perhaps the most promising models

va ehtimol eng optimal model

are those that mimic the relationship between neurons in the brain.

miyadagi neyronlar orasidagi bog’lanishni o’xshatishdir.

These artificial neural networks can use millions of connections

Bu sun’iy neyron bog’lamalari millionlab bog’lanishlarni

to tackle difficult tasks like image recognition, speech recognition,

qiyin muammolarni yechishda ishlata oladilar: surat, ovoz aniqlash

and even language translation.

va hattoki til tarjima qilish.

However, the more self-directed these models become,

Biroq, qanchalik o’zia yo’nalgan model bo’lsa,

the harder it is for computer scientists

dasturchilar uchun shunchalik bu

to determine how these self-taught algorithms arrive at their solution.

o’zi o’rgatilgan algoritmlar ishlashini aniqlash qiyin bo’ladi.

Researchers are already looking at ways to make machine learning more transparent.

Tadqiqotchilar o’zi o’rganadigan mashina- lar yaratishni yo’llarini ko’rmoqdalar.

But as AI becomes more involved in our everyday lives,

Biroq SI ko’proq hayotimizga aralashar ekan,

these enigmatic decisions have increasingly large impacts

bunday o’ylantiradigan qarorlarning ko’proq

on our work, health, and safety.

ish, sog’liq va xavfsizligimizga ta’siri bor.

So as machines continue learning to investigate, negotiate and communicate,

Shunday qilib mashinalar qidirishni, kelishish va gaplashishni o’rganar ekan,

we must also consider how to teach them to teach each other to operate ethically.

biz ularning bir-biriga to’g’ri o’rgatishini ham hisobga olishimiz kerak.