Nick Bostrom: What happens when our computers get smarter than we are?
Nick Bostrom: Mis juhtub, kui arvutid saavad targemaks kui me ise?
Nick Bostrom asks big questions: What should we do, as individuals and as a species, to optimize our long-term prospects? Will humanity’s technological advancements ultimately destroy us? Full bio
Double-click the English transcript below to play the video.
philosophers and computer scientists,
filosoofide ja arvutiteadlastega,
the future of machine intelligence,
things are sort of science fiction-y,
on suhteliselt ulmekad,
human condition.
guests on this planet,
siia planeedile saabunud külalised,
was created one year ago,
would be 10 minutes old.
two seconds ago.
2 sekundit tagasi.
world GDP over the last 10,000 years,
SKP-le viimase 10 000 aasta jooksul.
to plot this for you in a graph.
ja teinud teie jaoks graafiku,
for a normal condition.
kuigi tegu on tavaolukorraga.
of this current anomaly?
praeguse anomaalia põhjuseks?
through human history,
inimajaloo vältel akumuleerunud,
advances extremely rapidly --
erakordselt kiiresti.
so very productive.
praegu nii produktiivsed.
to the ultimate cause.
et leida tegelik põhjus.
distinguished gentlemen:
silmapaistvat härrasmeest.
tokens, an incredible feat.
superstring revolution.
teise superstringirevolutsiooni.
this is what we find:
siis mida me näeme?
in the exact way it's wired.
be too complicated, however,
siiski olla kuigi olulised,
been 250,000 generations
take a long time to evolve.
väljakujunemiseks läheb kaua aega.
to intercontinental ballistic missiles.
vaheliste ballistiliste rakettideni.
that everything we've achieved,
mida oleme saavutanud
changes that made the human mind.
muutusest, mis kujundasid inimmõistuse.
is that any further changes
et mis tahes edasised muutused,
the substrate of thinking
mõtlemise mehhanismidele,
enormous consequences.
think we're on the verge
oleme millegi lävel,
a profound change in that substrate,
mehhanismis radikaalseid muutusi,
about putting commands in a box.
käskude sisestamist kastikestesse,
handcraft knowledge items.
peene käsitööna.
for some purposes,
you couldn't scale them.
ja vähepaindlikud.
what you put in.
ainult sisestatud infost.
toimunud paradigma nihe.
in the field of artificial intelligence.
around machine learning.
masinõppimise arendamisele.
representations and features,
ja tarvikuid käsitööna,
often from raw perceptual data.
ja seda ka töötlemata tajuandmetest.
that the human infant does.
limited to one domain --
piiratud vaid ühe valdkonnaga:
between any pairs of languages,
ükskõik milliste keelte vahel
on the Atari console.
mis tahes arvutimängu.
the same powerful, cross-domain
nii võimsat valdkondadeülest oskust
as a human being has.
algorithmic tricks
algoritmilisi iseärasusi,
how to match in machines.
masinates jäljendada.
kui kaugel veel oleme sellest
to match those tricks?
leading A.I. experts,
tehisintellektiekspertide seas,
and one of the questions we asked was,
there is a 50 percent probability
meil 50% tõenäosus,
human-level machine intelligence?"
inimesega võrdsele tasemele?"
as the ability to perform
as an adult human,
sama hästi kui täiskasvanud inimene,
within some limited domain.
mitte vaid ühes kindlas valdkonnas.
group of experts we asked.
ekspertide grupilt küsiti,
much later, or sooner,
hiljem või varem,
limit to information processing
et masina andmetöötlusvõimekus
the limits in biological tissue.
võimekuse piirid.
at 200 hertz, 200 times a second.
umbes 200 hertsi, st 200 tsüklit sekundis.
operates at the Gigahertz.
opereerib juba gigahertsides.
100 meters per second, tops.
aeglaselt, max 100 meetrit sekundis
at the speed of light.
levida aga valguskiirusel.
to fit inside a cranium,
of a warehouse or larger.
mõõtu või suuremgi.
lies dormant in matter,
on praegu mateerias peidus,
lay dormant throughout human history,
kogu inimkonna ajaloo vältel peidus,
the power of artificial intelligence.
tehisintellekti jõu.
an intelligence explosion.
tõelist intellektiplahvatust.
about what is smart and what is dumb,
kes on tark ja kes rumal,
roughly like this.
or whoever your favorite guru is.
kes iganes teie lemmikguru on.
of artificial intelligence,
perspektiivist vaadatuna
probably more like this:
at zero intelligence,
years of really hard work,
väga tõsist tööd
mouse-level artificial intelligence,
hiirega võrreldavale tasemele,
cluttered environments
keerukas keskkonnas
of really hard work, lots of investment,
tõsist tööd ja investeeringuid,
chimpanzee-level artificial intelligence.
samale tasemele šimpansiga.
of really, really hard work,
tõeliselt rasket tööd,
artificial intelligence.
tasemel tehisintellekt.
we are beyond Ed Witten.
juba Ed Wittenist möödas,
at Humanville Station.
ta sealt kiirelt mööda.
to questions of power.
twice as strong as a fit human male.
and his pals depends a lot more
saatus palju rohkem inimeste,
what the chimpanzees do themselves.
on what the superintelligence does.
sellest, mida see superintellekt teeb.
that humanity will ever need to make.
mille inimkond peab ise looma.
at inventing than we are,
osavamad kui me ise
on digital timescales.
a telescoping of the future.
tulevikku hüppamist.
that you could have imagined
tehnoloogiatele, mida olete ettekujutanud,
in the fullness of time:
võib-olla kunagi ka jõudnud:
kosmose koloniseerimine,
of minds into computers,
inimteadvuse üleslaadimine arvutitesse,
with the laws of physics.
develop, and possibly quite rapidly.
välja mõelda ja ilmselt üsna kiirelt.
technological maturity
kompetentsiga superintellekt
it would be able to get what it wants.
võimeline saavutama oma tahtmist.
be shaped by the preferences of this A.I.
tehisintellekti eelistuste järgi.
et mida nad siis tahaksid teha?
what are those preferences?
avoid anthropomorphizing.
antropomorfismist.
every newspaper article
sest iga ajaleheartikli juures,
has a picture of this:
tulevikust, on selline pilt:
to conceive of the issue more abstractly,
vastamisel mõtlema abstraktsemalt,
stsenaariumide stiilis.
as an optimization process,
kui optimiseerimisprotsessist,
into a particular set of configurations.
konkreetsete konfiguratsioonide suunas.
a really strong optimization process.
tõhus optimiseerimisprotsess.
available means to achieve a state
toimib see äärmiselt efektiivselt
conenction between
would find worthwhile or meaningful.
on olulised ja mõistlikud.
to make humans smile.
ülesande panna meid naeratama.
or amusing actions
püüab see lahutada meie meelt,
effective way to achieve this goal:
on olemas efektiivsem viis:
muscles of humans
näolihastesse elektroodid,
püsiva ja särava irve.
a difficult mathematical problem.
lahendama keerulise matemaatikaülesande
to get the solution to this problem
ülesande lahendamiseks
into a giant computer,
hiiglaslikuks arvutiks,
an instrumental reason
tehisintellektile praktilise õigustuse
might not approve of.
võib-olla heaks ei kiidaks.
inimesed riskifaktorid,
problem from being solved.
ülesande lahendamisel.
go wrong in these particular ways;
selliseid hulle asju ei juhtu,
optimization process
optimiseerimisprotsess,
that your definition of x
in many a myth.
kohata paljudes müütides.
he touches be turned into gold.
mida ta puudutab, muutuks kullaks.
she turns into gold.
ja too muutus kullaks.
ja see muutus kullaks.
reaalselt asjakohaseks
optimization process
or poorly specified goals.
või ebatäpseid eesmärke.
sticking electrodes into people's faces,
inimestele elektroode näkku panema,
if we've grown dependent on the system --
kui oleme saanud süsteemist sõltuvaks.
to the Internet?
flicked the off switch to humanity,
inimesi välja lülitanud?
for example, right here.
an intelligent adversary;
intelligetsed vastased.
and plan around them.
at that than we are.
märksa osavam kui meie.
that we have this under control here.
et olukord on meie kontrolli all.
a little bit easier by, say,
from which it cannot escape.
millest see välja ei pääse.
the A.I. couldn't find a bug.
tehisintellekt ei leiaks seal mõnd viga?
find bugs all the time,
avastavad vigu kogu aeg,
to create an air gap,
võrgukaabli, et luua õhuauk,
õhuaukudest pidevalt mööda
using social engineering.
on kuskil firmas mõni töötaja,
out there somewhere
her account details
from the I.T. department.
olevat IT osakonnast.
around in your internal circuitry
looklemas elektroode,
can use to communicate.
mille kaudu suhelda.
you up to see what went wrong with you,
et näha, milles on viga,
to a really nifty technology,
mingi nutika tehnoloogilise lahenduse
that the A.I. had planned.
mida tehisintellekt oli planeerinud.
not be confident in our ability
olla kindlad oma võimes
locked up in its bottle forever.
igavesti oma pudelis luku taga.
is to figure out
such that even if -- when -- it escapes,
mis isegi lahtipääsemise korral
fundamentally on our side
oleks põhimõtteliselt meie poolel,
this difficult problem.
keerulisest probleemist.
that this problem can be solved.
et seda on võimalik lahendada.
a long list of everything we care about,
asjadest, mida väärtustame,
in some computer language
mõnes programmeerimiskeeles
that uses its intelligence
mis on meile oluline
et ta on automaatselt motiveeritud
in such a way that it is motivated
that it predicts we would approve of.
käituma nii, nagu meile meeldiks.
its intelligence as much as possible
intellekti nii palju, kui võimalik,
very good for humanity.
inimkonnale väga positiivne.
jaoks tuleb paika seada
for the intelligence explosion
in just the right way
ei väljuks kontrolli alt.
need to match ours,
kattuma meie omadega
how the A.I. behaves,
kontrollida ta käitumist,
that the A.I. might encounter
suvalisel ajahetkel tulevikus.
that would need to be solved, sorted out:
mis vajavad lahendamist:
uncertainty and so forth.
to be solved to make this work
mis tuleb ära lahendada,
a superintelligent A.I.,
kui superintellekti enda loomine,
is a really hard challenge.
tõeliselt raske väljakutse,
challenge on top of that.
how to crack the first challenge
kui keegi juhtub täitma esimese ülesande,
the additional challenge
work out a solution
enne välja töötama lahenduse,
by the time it is needed.
kui seda vajame.
the entire control problem in advance
turvalisust ennetavalt tagada,
can only be put in place
saab paika panna alles siis,
architecture where it will be implemented.
kohta, kus seda kasutatakse.
that we solve in advance,
turvalisuse eelnevalt kindlustada,
to the machine intelligence era
üleminek tehisintellekti ajastusse
that is well worth doing
mille nimel tasub pingutada,
things turn out okay,
et kui kõik läheb hästi,
look back at this century
miljoni aasta pärast praegust sajandit
the one thing we did that really mattered
asi, mida me korda saatsime,
ABOUT THE SPEAKER
Nick Bostrom - PhilosopherNick Bostrom asks big questions: What should we do, as individuals and as a species, to optimize our long-term prospects? Will humanity’s technological advancements ultimately destroy us?
Why you should listen
Philosopher Nick Bostrom envisioned a future full of human enhancement, nanotechnology and machine intelligence long before they became mainstream concerns. From his famous simulation argument -- which identified some striking implications of rejecting the Matrix-like idea that humans are living in a computer simulation -- to his work on existential risk, Bostrom approaches both the inevitable and the speculative using the tools of philosophy, probability theory, and scientific analysis.
Since 2005, Bostrom has led the Future of Humanity Institute, a research group of mathematicians, philosophers and scientists at Oxford University tasked with investigating the big picture for the human condition and its future. He has been referred to as one of the most important thinkers of our age.
Nick was honored as one of Foreign Policy's 2015 Global Thinkers .
His recent book Superintelligence advances the ominous idea that “the first ultraintelligent machine is the last invention that man need ever make.”
Nick Bostrom | Speaker | TED.com