How does artificial intelligence learn? - Briana Brownell

648,440 views ・ 2021-03-11

TED-Ed


Please double-click on the English subtitles below to play the video.

Prevodilac: Simonida Jekic Lektor: Milenka Okuka
00:09
Today, artificial intelligence helps doctors diagnose patients,
0
9829
5167
Danas veštačka inteligencija pomaže doktorima da dijagnostikuju pacijente,
00:14
pilots fly commercial aircraft, and city planners predict traffic.
1
14996
5042
pilotima da lete komercijalnim avionima i urbanistima da predvide saobraćaj.
00:20
But no matter what these AIs are doing, the computer scientists who designed them
2
20038
4250
Bez obzira na to šta ova VI radila, informatičari koji su je dizajnirali
00:24
likely don’t know exactly how they’re doing it.
3
24288
2750
verovatno ne znaju kako ona tačno radi.
00:27
This is because artificial intelligence is often self-taught,
4
27038
3875
To je iz razloga što je veštačka inteligencija najčešće samouka
00:30
working off a simple set of instructions
5
30913
2374
i radi na osnovu jednostavne grupe komandi
00:33
to create a unique array of rules and strategies.
6
33287
3584
da bi stvorila jedinstveni niz pravila i strategija.
00:36
So how exactly does a machine learn?
7
36871
2625
Kako tačno mašina uči?
00:39
There are many different ways to build self-teaching programs.
8
39496
3000
Postoje različiti načini kako se prave samouki programi.
00:42
But they all rely on the three basic types of machine learning:
9
42496
3916
Ali su svi zasnovani na tri osnovne vrste mašinskog učenja:
00:46
unsupervised learning, supervised learning, and reinforcement learning.
10
46412
5000
nenadgledano učenje, nadgledano učenje i učenje potkrepljivanjem.
00:51
To see these in action,
11
51412
1959
Da bismo ih videli na delu,
00:53
let’s imagine researchers are trying to pull information
12
53371
3250
zamislimo istraživače koji pokušavaju da izvuku informacije
00:56
from a set of medical data containing thousands of patient profiles.
13
56621
4208
iz grupe medicinskih podataka koji sadrže hiljade profila pacijenata.
01:01
First up, unsupervised learning.
14
61371
3125
Prvo imamo nenadgledano učenje.
01:04
This approach would be ideal for analyzing all the profiles
15
64496
3333
Ovaj pristup bi bio idealan za analiziranje svih profila
01:07
to find general similarities and useful patterns.
16
67829
3625
i pronalaženje sličnosti i korisnih obrazaca.
01:11
Maybe certain patients have similar disease presentations,
17
71454
3417
Možda pojedini pacijenti imaju slične kliničke slike
01:14
or perhaps a treatment produces specific sets of side effects.
18
74871
4042
ili terapija ima određeni skup neželjenih dejstava.
01:18
This broad pattern-seeking approach can be used to identify similarities
19
78913
4125
Ovaj opšti pristup traženju obrazaca može se koristiti da se pronađu sličnosti
01:23
between patient profiles and find emerging patterns,
20
83038
3375
između profila pacijenata, kao i za pronalaženje novih obrazaca,
01:26
all without human guidance.
21
86413
2291
sve to bez ljudskog delovanja.
01:28
But let's imagine doctors are looking for something more specific.
22
88704
3209
Recimo da lekari traže nešto specifičnije.
01:32
These physicians want to create an algorithm
23
92371
2208
Ti lekari žele da stvore algoritam
01:34
for diagnosing a particular condition.
24
94579
2709
za dijagnostikovanje određene bolesti.
01:37
They begin by collecting two sets of data—
25
97288
2583
Počnu tako što sakupljaju dve grupe podataka -
01:39
medical images and test results from both healthy patients
26
99871
3375
medicinske slike i rezultate testova zdravih pacijenata,
01:43
and those diagnosed with the condition.
27
103246
2417
ali i od obolelih.
01:45
Then, they input this data into a program
28
105663
2583
Zatim te podatke unose u program
01:48
designed to identify features shared by the sick patients
29
108246
3375
dizajniran da prepozna karakteristike prisutne kod bolesnih pacijenata,
01:51
but not the healthy patients.
30
111621
2125
ali ne i zdravih.
01:53
Based on how frequently it sees certain features,
31
113746
3083
Na osnovu toga koliko često uoči određene karakteristike,
01:56
the program will assign values to those features’ diagnostic significance,
32
116829
4042
program će odrediti vrednosti koliko je neka karakteristika važna,
02:00
generating an algorithm for diagnosing future patients.
33
120871
3708
stvarajući algoritam za dijagnostikovanje budućih pacijenata.
02:04
However, unlike unsupervised learning,
34
124579
3167
Ali za razliku od nenadgledanog učenja,
02:07
doctors and computer scientists have an active role in what happens next.
35
127746
4625
doktori i informatičari imaju aktivnu ulogu u onome što sledi.
02:12
Doctors will make the final diagnosis
36
132371
2083
Doktori će napraviti konačnu dijagnozu
02:14
and check the accuracy of the algorithm’s prediction.
37
134454
2917
i proveriti preciznost predviđanja algoritma.
02:17
Then computer scientists can use the updated datasets
38
137871
2917
Onda informatičari mogu da koriste dopunjene setove podataka
02:20
to adjust the program’s parameters and improve its accuracy.
39
140788
3500
da prilagode parametre programa i poboljšaju njegovu preciznost.
02:24
This hands-on approach is called supervised learning.
40
144663
3166
Ovaj direktni pristup se naziva nadgledano učenje.
02:27
Now, let’s say these doctors want to design another algorithm
41
147829
3167
Recimo da ovi doktori žele da naprave još jedan algoritam
02:30
to recommend treatment plans.
42
150996
1750
za preporučivanje planova za lečenje.
02:32
Since these plans will be implemented in stages,
43
152746
2833
S obzirom na to da će ovi planovi biti primenjeni u fazama,
02:35
and they may change depending on each individual's response to treatments,
44
155579
3959
i možda će se menjati na osnovu reakcije pojedinaca na terapiju,
02:39
the doctors decide to use reinforcement learning.
45
159538
2833
doktori su odlučili da koriste učenje potkrepljivanjem.
02:42
This program uses an iterative approach to gather feedback
46
162746
3167
Ovaj program koristi učestan pristup za sakupljanje povratnih informacija
02:45
about which medications, dosages and treatments are most effective.
47
165913
4583
o tome koji lekovi, doze i terapije su najefikasniji.
02:50
Then, it compares that data against each patient’s profile
48
170496
3000
Zatim poredi podatke sa profilom svakog pacijenta
02:53
to create their unique, optimal treatment plan.
49
173496
2833
i stvara njihov jedinstveni, optimalni plan lečenja.
02:56
As the treatments progress and the program receives more feedback,
50
176329
3500
Kako lečenje napreduje i program dobija sve više povratnih informacija,
02:59
it can constantly update the plan for each patient.
51
179829
3292
on može stalno da ažurira plan za svakog pacijenta.
03:03
None of these three techniques are inherently smarter than any other.
52
183121
3375
Nijedna od ove tri tehnike nije “pametnija” od ostalih.
03:06
While some require more or less human intervention,
53
186496
2667
Nekima je potrebno više ili manje ljudske intervencije,
03:09
they all have their own strengths and weaknesses
54
189163
2416
sve imaju svoje prednosti i mane
03:11
which makes them best suited for certain tasks.
55
191579
2250
što ih čini odgovarajućim za određene zadatke.
03:14
However, by using them together,
56
194329
2375
No koristeći ih zajedno,
03:16
researchers can build complex AI systems,
57
196704
2875
istraživači su izgradili kompleksne sisteme veštačke inteligencije
03:19
where individual programs can supervise and teach each other.
58
199579
3292
gde pojedinačni programi mogu da nadgledaju i uče jedni druge.
03:22
For example, when our unsupervised learning program
59
202871
2958
Na primer, kada naš program nenadgledanog učenja
03:25
finds groups of patients that are similar,
60
205829
2334
pronađe grupu pacijenata koji su slični,
03:28
it could send that data to a connected supervised learning program.
61
208163
3375
može da pošalje te podatke u program nadgledanog učenja sa kojim je povezan.
03:31
That program could then incorporate this information into its predictions.
62
211829
3500
Taj program onda može da pripoji ovu informaciju u svoja predviđanja.
03:35
Or perhaps dozens of reinforcement learning programs
63
215871
2792
Ili desetine programa učenja potkrepljivanjem
03:38
might simulate potential patient outcomes
64
218663
2291
mogu da simuliraju moguće ishode kod pacijenata
03:40
to collect feedback about different treatment plans.
65
220954
2750
kako bi sakupili povratne informacije o različitim terapijama.
03:43
There are numerous ways to create these machine-learning systems,
66
223704
3125
Postoje mnogi načini da se naprave ovakvi sistemi mašinskog učenja,
03:46
and perhaps the most promising models
67
226829
1834
a možda najobećavajući
03:48
are those that mimic the relationship between neurons in the brain.
68
228663
3416
su oni koji imitiraju vezu između neurona u mozgu.
03:52
These artificial neural networks can use millions of connections
69
232079
3292
Ove veštačke neuronske mreže mogu da koriste milione veza
03:55
to tackle difficult tasks like image recognition, speech recognition,
70
235371
4417
za teške zadatke poput prepoznavanja slika, govora,
03:59
and even language translation.
71
239788
2041
pa čak i prevođenja sa drugih jezika.
04:01
However, the more self-directed these models become,
72
241829
3292
Ali što su ovi modeli samostalniji,
04:05
the harder it is for computer scientists
73
245121
2125
teže je informatičarima
04:07
to determine how these self-taught algorithms arrive at their solution.
74
247246
3833
da odrede kako samouki algoritmi dolaze do svojih zaključaka.
04:11
Researchers are already looking at ways to make machine learning more transparent.
75
251079
4459
Istraživači već traže načine da učine mašinsko učenje transparentnijim.
04:15
But as AI becomes more involved in our everyday lives,
76
255538
2916
Međutim, kako VI postaje sve više uključena u naš svakodnevni život,
04:18
these enigmatic decisions have increasingly large impacts
77
258454
2792
ove enigmatične odluke imaju mnogo veći uticaj
04:21
on our work, health, and safety.
78
261246
2875
na naše poslove, zdravlje i bezbednost.
04:24
So as machines continue learning to investigate, negotiate and communicate,
79
264121
4958
Kako mašine nastavljaju da uče da istražuju, pregovaraju i komuniciraju
04:29
we must also consider how to teach them to teach each other to operate ethically.
80
269079
5209
moramo takođe uzeti u obzir da ih naučimo da uče jedna drugu da rade etički.
About this website

This site will introduce you to YouTube videos that are useful for learning English. You will see English lessons taught by top-notch teachers from around the world. Double-click on the English subtitles displayed on each video page to play the video from there. The subtitles scroll in sync with the video playback. If you have any comments or requests, please contact us using this contact form.

https://forms.gle/WvT1wiN1qDtmnspy7