The Turing test: Can a computer pass for a human? - Alex Gendler

1,661,434 views ・ 2016-04-25

TED-Ed


Dvaput kliknite na engleske titlove ispod za reprodukciju videozapisa.

Prevoditelj: Katarina Mikjel Recezent: Sanda L
00:06
What is consciousness?
0
6999
2035
Što je svijest?
00:09
Can an artificial machine really think?
1
9034
2794
Može li umjetni stroj zaista misliti?
00:11
Does the mind just consist of neurons in the brain,
2
11828
3286
Čine li um samo neuroni u mozgu
00:15
or is there some intangible spark at its core?
3
15114
3643
ili je u njegovoj srži neka neopipljiva iskra?
00:18
For many, these have been vital considerations
4
18757
2568
Po mišljenju mnogih, to su ključna pitanja
00:21
for the future of artificial intelligence.
5
21325
3027
za budućnost umjetne inteligencije.
00:24
But British computer scientist Alan Turing decided to disregard all these questions
6
24352
5187
Britanski računalni znanstvenik Alan Turing zanemario ih je
00:29
in favor of a much simpler one:
7
29539
2311
i postavio jedno jednostavnije:
00:31
can a computer talk like a human?
8
31850
3419
može li računalo razgovarati kao čovjek?
00:35
This question led to an idea for measuring aritificial intelligence
9
35269
3996
To je pitanje dovelo do zamisli o mjerenju umjetne inteligencije,
00:39
that would famously come to be known as the Turing test.
10
39265
4051
koja će postati poznata kao Turingov test.
00:43
In the 1950 paper, "Computing Machinery and Intelligence,"
11
43316
4025
U znanstvenom radu „Strojevi za računanje i inteligencija” iz 1950.
00:47
Turing proposed the following game.
12
47341
2473
Turing je predložio sljedeću igru.
00:49
A human judge has a text conversation with unseen players
13
49814
4092
Sudac (čovjek) dopisuje se s igračima koje ne vidi
00:53
and evaluates their responses.
14
53906
2476
i ocjenjuje njihove odgovore.
00:56
To pass the test, a computer must be able to replace one of the players
15
56382
4027
Da bi prošlo test, računalo mora zamijeniti nekog od igrača,
01:00
without substantially changing the results.
16
60409
3448
a da pritom ne dođe do značajne promjene u rezultatima.
01:03
In other words, a computer would be considered intelligent
17
63857
2959
Drugim riječima, računalo bismo smatrali inteligentnim
01:06
if its conversation couldn't be easily distinguished from a human's.
18
66816
5697
ako njegov govor nije lako razlikovati od ljudskog.
01:12
Turing predicted that by the year 2000,
19
72513
2154
Turing je predvidio da će do 2000. godine
01:14
machines with 100 megabytes of memory would be able to easily pass his test.
20
74667
5955
strojevi s memorijom od 100 megabajta uspjeti proći njegov test bez poteškoća.
01:20
But he may have jumped the gun.
21
80622
2164
Možda se ipak malo zaletio.
01:22
Even though today's computers have far more memory than that,
22
82786
3054
Iako današnja računala imaju znatno veću memoriju,
01:25
few have succeeded,
23
85840
1803
malo je njih u tome uspjelo,
01:27
and those that have done well
24
87643
1795
a oni koji jesu imali dobre rezultate,
01:29
focused more on finding clever ways to fool judges
25
89438
3770
više su bili usredotočeni na to kako pametno zavarati suce,
01:33
than using overwhelming computing power.
26
93208
2966
nego na to kako upotrijebiti ogromnu računalnu snagu.
01:36
Though it was never subjected to a real test,
27
96174
2922
Iako nikad nije podvrgnut stvarnom testu,
01:39
the first program with some claim to success was called ELIZA.
28
99096
4769
prvi program koji se može pohvaliti barem djelomičnim uspjehom zvao se ELIZA.
01:43
With only a fairly short and simple script,
29
103865
2547
Taj prilično kratak i jednostavan program
01:46
it managed to mislead many people by mimicking a psychologist,
30
106412
4022
uspio je obmanuti mnoge ljude oponašajući psihologa,
01:50
encouraging them to talk more
31
110434
1961
tako što ih je ohrabrivao da više govore
01:52
and reflecting their own questions back at them.
32
112395
3510
i uzvraćao im njihovim vlastitim pitanjima.
01:55
Another early script PARRY took the opposite approach
33
115905
3507
Drugi rani program PARRY postupio je suprotno,
01:59
by imitating a paranoid schizophrenic
34
119412
2801
oponašajući paranoičnog šizofreničara
02:02
who kept steering the conversation back to his own preprogrammed obsessions.
35
122213
5726
koji je stalno usmjeravao razgovor na vlastite pretprogramirane opsesije.
02:07
Their success in fooling people highlighted one weakness of the test.
36
127939
4864
Njihovi uspjesi u obmanjivanju ljudi razotkrili su slabu točku testa.
02:12
Humans regularly attribute intelligence to a whole range of things
37
132803
4630
Ljudi redovito pripisuju inteligenciju čitavom nizu pojava
02:17
that are not actually intelligent.
38
137433
3643
koje s njom zapravo nemaju veze.
02:21
Nonetheless, annual competitions like the Loebner Prize,
39
141076
3213
Unatoč tome, godišnja natjecanja, kao što je Loebnerova nagrada,
02:24
have made the test more formal
40
144289
1720
formalizirala su test.
02:26
with judges knowing ahead of time
41
146009
2146
Suci sad unaprijed znaju
02:28
that some of their conversation partners are machines.
42
148155
3864
da su neki od njihovih sugovornika strojevi.
02:32
But while the quality has improved,
43
152019
1901
No iako se kvaliteta poboljšala,
02:33
many chatbot programmers have used similar strategies to ELIZA and PARRY.
44
153920
5176
mnogi programeri chatbotâ koristili su se strategijama sličnima ELIZA-i i PARRY-ju.
02:39
1997's winner Catherine
45
159096
2036
Pobjednica iz 1997. Catherine
02:41
could carry on amazingly focused and intelligent conversation,
46
161132
4081
mogla je voditi zapanjujuće usredotočene i inteligentne razgovore,
02:45
but mostly if the judge wanted to talk about Bill Clinton.
47
165213
3962
ali uglavnom ako bi sudac želio razgovarati o Billu Clintonu.
02:49
And the more recent winner Eugene Goostman
48
169175
2638
A nedavnom pobjedniku Eugeneu Goostmanu
02:51
was given the persona of a 13-year-old Ukrainian boy,
49
171813
3736
dana je osobnost trinaestogodišnjaka iz Ukrajine,
02:55
so judges interpreted its nonsequiturs and awkward grammar
50
175549
4021
pa su suci njegove nesuvisle rečenice i nespretnu gramatiku pogrešno protumačili
02:59
as language and culture barriers.
51
179570
3346
kao jezičnu i kulturološku barijeru.
03:02
Meanwhile, other programs like Cleverbot have taken a different approach
52
182916
4219
U međuvremenu su drugi programi, npr. Cleverbot, odabrali drukčiji pristup:
03:07
by statistically analyzing huge databases of real conversations
53
187135
4605
statističkom analizom golemih baza podataka stvarnih razgovora
03:11
to determine the best responses.
54
191740
2541
birali su najbolje odgovore.
03:14
Some also store memories of previous conversations
55
194281
3249
Neki su i pohranjivali u memoriji prijašnje razgovore
03:17
in order to improve over time.
56
197530
3385
kako bi se s vremenom poboljšali.
03:20
But while Cleverbot's individual responses can sound incredibly human,
57
200915
4151
No iako Cleverbotovi odgovori zasebno mogu zvučati nevjerojatno ljudski,
03:25
its lack of a consistent personality
58
205066
2026
nedostatak dosljedne osobnosti
03:27
and inability to deal with brand new topics
59
207092
3153
i nesposobnost prilagodbe sasvim novim temama
03:30
are a dead giveaway.
60
210245
2613
smjesta ga odaju.
03:32
Who in Turing's day could have predicted that today's computers
61
212858
3254
Tko je u Turingovo vrijeme mogao predvidjeti da će današnja računala
03:36
would be able to pilot spacecraft,
62
216112
2089
moći upravljati svemirskim letjelicama,
03:38
perform delicate surgeries,
63
218201
2317
obavljati osjetljive kirurške operacije
03:40
and solve massive equations,
64
220518
2289
i rješavati složene jednadžbe,
03:42
but still struggle with the most basic small talk?
65
222807
3503
no i dalje zapinjati u običnom čavrljanju?
03:46
Human language turns out to be an amazingly complex phenomenon
66
226310
3509
Ljudski se jezik pokazao nevjerojatno složenom pojavom
03:49
that can't be captured by even the largest dictionary.
67
229819
3705
koju ne može obuhvatiti čak ni najveći rječnik.
03:53
Chatbots can be baffled by simple pauses, like "umm..."
68
233524
4469
Chatbote zbunjuju jednostavne pauze, npr. „hm...”
03:57
or questions with no correct answer.
69
237993
2422
ili pitanja bez točnih odgovora.
04:00
And a simple conversational sentence,
70
240415
1900
A za analizu rečenice najbanalnijeg razgovora,
04:02
like, "I took the juice out of the fridge and gave it to him,
71
242315
3280
primjerice „Izvadio sam sok iz hladnjaka i dao mu ga,
04:05
but forgot to check the date,"
72
245595
1811
ali sam zaboravio provjeriti datum.”,
04:07
requires a wealth of underlying knowledge and intuition to parse.
73
247406
5466
potrebni su ogromno prethodno znanje i intuicija.
04:12
It turns out that simulating a human conversation
74
252872
2631
I tako se ispostavilo da je za simulaciju ljudskog razgovora
04:15
takes more than just increasing memory and processing power,
75
255503
3609
potrebno više od pukog povećanja memorije i snage procesora
04:19
and as we get closer to Turing's goal,
76
259112
2511
i da ćemo se na putu prema Turingovu cilju
04:21
we may have to deal with all those big questions about consciousness after all.
77
261623
4860
ipak trebati pozabaviti svim tim velikim pitanjima o svijesti.
O ovoj web stranici

Ova stranica će vas upoznati s YouTube videozapisima koji su korisni za učenje engleskog jezika. Vidjet ćete lekcije engleskog koje vode vrhunski profesori iz cijelog svijeta. Dvaput kliknite na engleske titlove prikazane na svakoj video stranici da biste reproducirali video s tog mjesta. Titlovi se pomiču sinkronizirano s reprodukcijom videozapisa. Ako imate bilo kakvih komentara ili zahtjeva, obratite nam se putem ovog obrasca za kontakt.

https://forms.gle/WvT1wiN1qDtmnspy7