Blaise Agüera y Arcas: How computers are learning to be creative
Blaise Agüera y Arcas: Bilgisayarlar yaratıcı olmayı nasıl öğrenirler
Blaise Agüera y Arcas works on machine learning at Google. Previously a Distinguished Engineer at Microsoft, he has worked on augmented reality, mapping, wearable computing and natural user interfaces. Full bio
Double-click the English transcript below to play the video.
that works on machine intelligence;
bir takımı yönetiyorum;
of making computers and devices
makineleri yapan mühendislik disiplini
that brains do.
interested in real brains
in the things that our brains do
performansından hâlâ üstün olan
to the performance of computers.
ilgi duymamızı sağlıyor.
has been perception,
alanlardan biri algıdır,
out there in the world --
kullanışlıdır.
for example, that our team makes,
algı algoritmaları örneğin,
on Google Photos to become searchable,
şeyleri baz alarak onların aranabilir
out there into the world.
dönüştürmektir.
our work on machine perception
yaptığımız çalışmalar da
with the world of machine creativity
yaratıcılığı ve makine sanatı dünyasıyla
had a penetrating insight
arasındaki ikili ilişkiye
between perception and creativity.
has a statue inside of it,
barındırır ve
is to discover it."
çıkarmaktır."
Michelangelo was getting at
Michelangelo'nun anlatmak istediği
is an act of imagination
bir eylemi ve
and perceiving and imagining,
sağlayan organımız
with a brief bit of history
the heart or the intestines,
about a brain by just looking at it,
bakarak bir beyin hakkında
of this thing all kinds of fanciful names,
anlamına gelen hipokampüs gibi
doesn't tell us very much
developed some kind of insight
bir bakış açısı geliştiren ilk kişi,
Santiago Ramón y Cajal,
başlamak için
or render in very high contrast
eritebilen özel kimyasallar kullanan,
their morphologies.
that he made of neurons
(sinir hücresi) yaptığı
of different sorts of cells,
hücreleri görüyorsunuz,
was quite new at this point.
very, very long distances --
to some people in the 19th century;
için açık ve netti;
were just getting underway.
yeni bir yolculuğa başlıyordu.
of Ramón y Cajal's, like this one,
mikroanatomik çizimleri
that Ramón y Cajal started.
sonrasından daha uzağız.
Enstitüsü'ndeki
of Neuroscience.
hayal etmekti.
is about one cubic millimeter in size,
bir milimetre kübü ve
very small piece of it here.
bir parçasını gösteriyorum.
tiny block of tissue.
ard arda gelen dilimler.
of hair is about 100 microns.
yaklaşık 100 mikron.
much, much smaller
electron microscopy slices,
elektron mikroskopi dilimlerinden
in 3D of neurons that look like these.
yeniden yapılandırmaya başlayabilir.
style as Ramón y Cajal.
aynı tarzda.
be able to see anything here.
one neuron to another.
ahead of his time,
ilerisindeydi
over the next few decades.
biliyorduk ve
was advanced enough
çalıştıklarını daha iyi
experiments on live neurons
elektrikli deneyleri
when computers were being invented,
bilgisayar biliminin fikir
of modeling the brain --
Alan Turing'in deyimiyle
as Alan Turing called it,
"akıllı makinelerin"
Ramón y Cajal'ın burada gösterdiğim
looked at Ramón y Cajal's drawing
imagery that comes from the eye.
görüntüleri işliyor.
like a circuit diagram.
in McCulloch and Pitts's circuit diagram
devre şemasında
birçok detay bulunmakta.
of computational elements
elemanları serisi gibi
one to the next in a cascade,
ard arda bilgi aktarması fikri
visual information would need to do.
for us to do with our brains.
yaptığımız çok kolay bir şeydir.
that for a computer,
bilgisayar için,
just a few years ago.
this task is easy to do.
biri değil.
and the word "bird,"
arasında olan,
connected to each other
inside our visual cortices,
içerisinde, biyolojik olabilir
to have the capability
sinir ağları modelleme
on the computer.
başladık bilgisayarlarda.
göründüğünü açıklayacağım.
that actually looks like.
about as a first layer of neurons,
katmanı gibi düşünebilirsiniz
how it works in the eye --
bu şekilde işler --
after another layer of neurons,
ardından diğer farklı ağırlıklı
of different weights.
olan nöronlar katmanına.
of all of those synapses.
karakterize edilir.
properties of this network.
karakterize ederler.
or a small group of neurons
nöron grubunuz olur,
those three things --
in the neural network,
sinir ağındaki sinapslar
ya da civarında x --
these synapses in the neural network.
ağırlıklarını gösteren.
is just a simple formula,
farz edelim:
arasına aldım çünkü
going on there, of course,
of mathematical operations.
çok karmaşık serileridir.
that if you have one equation,
eğer bir denklemin varsa,
by knowing the other two things.
bulabilirsin.
that the picture of a bird is a bird,
and w and x are known.
you know the pixels.
pikselleri biliyorsunuz.
a relatively straightforward problem.
anlaşılır bir problem.
and you're done.
aynısını yapan
doing exactly that.
on a mobile phone,
cep telefonunda işletiliyor
amazing in its own right,
billions and trillions of operations
milyarlarca ve trilyonlarca işlemi
picture of a bird,
"Yes, it's a bird,"
demekle kalmıyor
with a network of this sort.
cinsini de tanımlıyor.
and the y is the unknown.
y bilinmiyor.
difficult part, of course,
do we figure out the w,
of solving for w,
hesaplama sürecinde
with the simple equation
it's the inverse to multiplication,
çarpmanın tersi,
very non-linear operation;
olmayan bir işlem;
to solve the equation
bir yol bulmalıyız
is fairly straightforward.
a little algebra trick,
to the right-hand side of the equation.
sağ tarafına taşıyalım.
about it as an error.
for w the right way,
to minimize the error,
tahminlerde bulunabiliriz
computers are very good at.
olduğu bir alan.
Hata 4 olur.
sort of play Marco Polo,
körebe (Marco Polo)
0'a yakınlaştırabilir.
successive approximations to w.
için ardışık yaklaşıklama elde eder.
but after about a dozen steps,
ama bir düzine aşamadan sonra
which is close enough.
ki bu da yeterince yakın.
a lot of known x's and known y's
through an iterative process.
süreç boyunca çözüyoruz.
that we do our own learning.
tamamen aynı.
this is not a bird."
for those neural connections.
çözüyoruz.
x and w fixed to solve for y;
sabit x ve w'ye sahibiz
Alex Mordvintsev, on our team,
ekibimizden Alex Mordvintsev
with what happens if we try solving for x,
çözmeye çalışırsak ne olacağını
that you've trained on birds,
zaten bulunmakta,
the same error-minimization procedure,
kuşları tanımlamak
trained to recognize birds,
yapılabileceği sonucu ortaya çıkar,
generated entirely by a neural network
eğitilen sinir ağı
rather than solving for y,
y'yi hesaplamak yerine
by Mike Tyka in our group,
tarafından yapılan bir çalışma,
of William Kentridge's artworks,
çalışmalarını hatırlatıyor,
over the space of different animals,
boşluklarda y'yi değiştiriyor,
ayırmak ve onları
to recognize and distinguish
diğerine biçimler elde edilir.
morph from one animal to another.
have tried reducing
out of the space of all things
over that entire surface,
you make a kind of map --
bir çeşit harita yapıyorsunuz,
the network knows how to recognize.
her şeyin görsel bir haritası.
"armadillo" is right in that spot.
"armadillo" tam bu noktada.
of networks as well.
to recognize faces,
için tasarlanmış,
in a y that says, "me,"
psychedelic picture of me
psikedelik resmimi üretiyor,
multiple points of view at once
sebebi ise ağın,
to get rid of the ambiguity
pozda ya da şu veya
or another pose,
olmasındaki anlaşmazlıktan
another kind of lighting.
this sort of reconstruction,
of different points of view,
bir tür bozulma elde edersiniz,
his own face as a guide image
Alex kendi yüzünü
to reconstruct my own face.
kullandığında böyle oluyor.
that optimization process.
gereken çok iş var.
more like a coherent face,
erişmeye başlıyorsunuz yorumlamada
with a blank canvas
that is itself already some other image.
ile başlayabilirsiniz.
that is designed to categorize
farklı nesneleri kategorize etmek
man-made structures, animals ...
insan yapımı yapılar, hayvanlar gibi...
with just a picture of clouds,
resmi ile başlıyoruz
what it sees in the clouds.
bulutlarda gördüklerini çözüyor.
you spend looking at this,
uzun süre harcarsanız,
will see in the clouds.
to hallucinate into this,
yüz ağını kullanabilir
bulandırıyor, yakınlaştırıyor,
zooms hallucinates, zooms.
bulandırıyor, yakınlaştırıyor.
of the network, I suppose,
edersiniz sanırım
is eating its own tail.
zemin oluşturmakta,
yapmayı düşünüyorum?"
What do I think I see next?"
called "Higher Education" --
adlı derste gösterdim
marijuana was legalized.
yasallaştırılmasından sonraydı.
is not constrained.
not ederek.
because they're really fun to look at.
çünkü bakması gerçekten eğlenceli.
a camera that takes a picture,
deneyler yaptı,
writes a poem using neural networks,
sırtında sinir ağı kullanarak
has been trained
are very intimately connected.
derinlemesine bağlıdır.
tanılamak için
things in the world,
Michelangelo really did see
içindeki bloklarda
any being, any alien
perceptual acts of that sort
ya da uzaylı da
machinery that's used in both cases.
tamamen aynı düzenek.
and creativity are by no means
ve yaratıcılık kesinlikle
that can do exactly these sorts of things.
modellerine sahip olmaya başladık.
the brain is computational.
beyin hesaba dayalıdır.
in designing intelligent machinery.
tasarlama alıştırmaları olarak başladı.
of those early pioneers,
is not just about accounting
sadece hesaplama
we modeled them after our minds.
onları zihnimize benzer modelledik.
to understand our own minds better
zihinlerimizi anlama, hem de
ABOUT THE SPEAKER
Blaise Agüera y Arcas - Software architectBlaise Agüera y Arcas works on machine learning at Google. Previously a Distinguished Engineer at Microsoft, he has worked on augmented reality, mapping, wearable computing and natural user interfaces.
Why you should listen
Blaise Agüera y Arcas is principal scientist at Google, where he leads a team working on machine intelligence for mobile devices. His group works extensively with deep neural nets for machine perception and distributed learning, and it also investigates so-called "connectomics" research, assessing maps of connections within the brain.
Agüera y Arcas' background is as multidimensional as the visions he helps create. In the 1990s, he authored patents on both video compression and 3D visualization techniques, and in 2001, he made an influential computational discovery that cast doubt on Gutenberg's role as the father of movable type.
He also created Seadragon (acquired by Microsoft in 2006), the visualization technology that gives Photosynth its amazingly smooth digital rendering and zoom capabilities. Photosynth itself is a vastly powerful piece of software capable of taking a wide variety of images, analyzing them for similarities, and grafting them together into an interactive three-dimensional space. This seamless patchwork of images can be viewed via multiple angles and magnifications, allowing us to look around corners or “fly” in for a (much) closer look. Simply put, it could utterly transform the way we experience digital images.
He joined Microsoft when Seadragon was acquired by Live Labs in 2006. Shortly after the acquisition of Seadragon, Agüera y Arcas directed his team in a collaboration with Microsoft Research and the University of Washington, leading to the first public previews of Photosynth several months later. His TED Talk on Seadragon and Photosynth in 2007 is rated one of TED's "most jaw-dropping." He returned to TED in 2010 to demo Bing’s augmented reality maps.
Fun fact: According to the author, Agüera y Arcas is the inspiration for the character Elgin in the 2012 best-selling novel Where'd You Go, Bernadette?
Blaise Agüera y Arcas | Speaker | TED.com