Blaise Agüera y Arcas: How computers are learning to be creative
Блез Агюера и Аркас: Как компьютеры учатся творчеству
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;
работающую над машинным интеллектом,
of making computers and devices
созданием компьютеров и устройств,
that brains do.
interested in real brains
настоящей мозговой деятельностью,
in the things that our brains do
нашего мозга,
to the performance of computers.
возможности компьютеров.
has been perception,
было восприятие —
out there in the world --
для нашего разума,
for example, that our team makes,
машинные алгоритмы восприятия,
on Google Photos to become searchable,
картинки на Google Photos по тому,
out there into the world.
our work on machine perception
наша работа над компьютерным восприятием
with the world of machine creativity
с миром компьютерного творчества
had a penetrating insight
было точное понимание
between perception and creativity.
между восприятием и творчеством.
has a statue inside of it,
is to discover it."
Michelangelo was getting at
is an act of imagination
and perceiving and imagining,
воспринимает и представляет, —
with a brief bit of history
с небольшого экскурса в историю
the heart or the intestines,
от сердца или кишечника,
about a brain by just looking at it,
просто посмотрев на него
of this thing all kinds of fanciful names,
этого óргана причудливые имена,
что означает «креветка».
doesn't tell us very much
мало говорят нам
developed some kind of insight
было некоторое понимание того,
Santiago Ramón y Cajal,
Сантьяго Рамон-и-Кахаль,
и специальный краситель,
or render in very high contrast
отдельные клетки мозга
their morphologies.
that he made of neurons
которые он нарисовал
of different sorts of cells,
различных видов клеток,
was quite new at this point.
была в то время довольно нова.
very, very long distances --
to some people in the 19th century;
для некоторых людей в XIX веке:
were just getting underway.
of Ramón y Cajal's, like this one,
Рамона-и-Кахаля
that Ramón y Cajal started.
начатую Рамоном-и-Кахалем.
от наших коллег
of Neuroscience.
ткани головного мозга.
is about one cubic millimeter in size,
один кубический миллиметр,
very small piece of it here.
очень маленький его кусочек.
это митохондрии,
tiny block of tissue.
of hair is about 100 microns.
около 100 микрон.
much, much smaller
electron microscopy slices,
in 3D of neurons that look like these.
нейронов, похожие на эти.
style as Ramón y Cajal.
на рисунки Рамона-и-Кахаля.
be able to see anything here.
one neuron to another.
один нейрон с другим.
ahead of his time,
своё время,
продвигался медленно
over the next few decades.
что нейроны используют электричество,
was advanced enough
технологии были развиты достаточно,
experiments on live neurons
эксперименты на живых нейронах,
when computers were being invented,
of modeling the brain --
моделирования мозга,
as Alan Turing called it,
как называл его Алан Тьюринг,
looked at Ramón y Cajal's drawing
посмотрели на рисунки Рамона-и-Кахаля.
imagery that comes from the eye.
образы, поступающие от органов зрения.
like a circuit diagram.
как замкнутая схема.
in McCulloch and Pitts's circuit diagram
многие детали
of computational elements
как серия вычислительных элементов,
one to the next in a cascade,
друг другу информацию,
visual information would need to do.
визуальной информации.
такую как эта, и сказать:
for us to do with our brains.
that for a computer,
что для компьютера
just a few years ago.
подобное было почти невозможно.
this task is easy to do.
которое можно выполнить легко.
and the word "bird,"
connected to each other
связана друг с другом
inside our visual cortices,
могла быть биологической,
to have the capability
on the computer.
на компьютере.
that actually looks like.
about as a first layer of neurons,
как о первом слое нейронов —
how it works in the eye --
after another layer of neurons,
of different weights.
синапсы различных весов.
of all of those synapses.
properties of this network.
свойства этой сети.
or a small group of neurons
those three things --
эти три вещи —
in the neural network,
в нейронной сети
these synapses in the neural network.
в нейронной сети.
is just a simple formula,
что это простая формула:
going on there, of course,
там просисходит
of mathematical operations.
математических расчётов.
that if you have one equation,
что если у нас есть уравнение,
by knowing the other two things.
зная две другие.
that the picture of a bird is a bird,
в изображении птицы
and w and x are known.
а W и X известны.
you know the pixels.
вы знаете пиксели.
a relatively straightforward problem.
простая задача.
and you're done.
doing exactly that.
делая то же самое.
on a mobile phone,
на мобильном телефоне.
amazing in its own right,
billions and trillions of operations
миллиарды и триллионы операций
picture of a bird,
одно за другим
"Yes, it's a bird,"
with a network of this sort.
определяет вид этой птицы.
and the y is the unknown.
а Y неизвестен.
difficult part, of course,
самую сложную часть —
do we figure out the w,
мы находим W —
of solving for w,
with the simple equation
как о числах,
6 = 2 * W,
it's the inverse to multiplication,
так как оно обратно умножению,
very non-linear operation;
нелинейная операция,
to solve the equation
решить это уравнение
is fairly straightforward.
a little algebra trick,
математический трюк
to the right-hand side of the equation.
about it as an error.
давайте примем его за отклонение.
for w the right way,
если мы посчитаем W правильно,
to minimize the error,
чтобы минимизировать отклонение,
computers are very good at.
в чём компьютеры очень хороши.
sort of play Marco Polo,
в «угадайку»
successive approximations to w.
but after about a dozen steps,
но сделав с десяток шагов,
which is close enough.
что достаточно близко к истине.
a lot of known x's and known y's
through an iterative process.
находим W.
that we do our own learning.
this is not a bird."
«Это птица, а это не птица».
for those neural connections.
x and w fixed to solve for y;
для нахождения Y
Alex Mordvintsev, on our team,
из нашей команды
with what happens if we try solving for x,
если попытаться найти X,
that you've trained on birds,
the same error-minimization procedure,
процедуру минимизации погрешностей,
trained to recognize birds,
обученной распознавать птиц,
generated entirely by a neural network
создано нейронной сетью,
rather than solving for y,
by Mike Tyka in our group,
из нашей команды.
of William Kentridge's artworks,
Уильяма Кентриджа,
потом их стирает,
over the space of different animals,
to recognize and distinguish
morph from one animal to another.
изображение животных.
have tried reducing
out of the space of all things
из пространства всех объектов,
over that entire surface,
на всей поверхности,
you make a kind of map --
вы делаете подобие карты —
the network knows how to recognize.
которые может распознать сеть.
"armadillo" is right in that spot.
Вот «броненосец».
of networks as well.
типами сетей.
to recognize faces,
in a y that says, "me,"
который значит «я»,
psychedelic picture of me
психоделическое изображение меня
multiple points of view at once
так выглядит,
to get rid of the ambiguity
от неопределённости,
or another pose,
another kind of lighting.
this sort of reconstruction,
of different points of view,
сделанных с разных сторон,
his own face as a guide image
своё лицо как опорное изображение
to reconstruct my own face.
по воссозданию моего лица.
that optimization process.
more like a coherent face,
with a blank canvas
that is itself already some other image.
является другим изображением.
that is designed to categorize
чтобы категоризировать
man-made structures, animals ...
искусственные сооружения, животных...
with just a picture of clouds,
с простого изображения облаков,
what it sees in the clouds.
что именно она видит в облаках.
you spend looking at this,
will see in the clouds.
to hallucinate into this,
чтобы вызвать галюцинации
zooms hallucinates, zooms.
галлюцинирует, приближает.
of the network, I suppose,
диссоциативной фуги сети
is eating its own tail.
для ответа на вопрос:
What do I think I see next?"
Что же я увижу дальше?»
called "Higher Education" --
marijuana was legalized.
легализации марихуаны.
is not constrained.
этой технологии безграничны.
because they're really fun to look at.
потому что на них интересно смотреть.
a camera that takes a picture,
на которую делал снимок,
writes a poem using neural networks,
пишущий стихи посредством нейронных сетей,
has been trained
была «натренирована»
are very intimately connected.
очень тесно связаны между собой.
things in the world,
воспроизводя их.
Michelangelo really did see
any being, any alien
даже инопланетное,
perceptual acts of that sort
machinery that's used in both cases.
одинаковые.
and creativity are by no means
что восприятие и творчество присущи
that can do exactly these sorts of things.
которые могут делать то же самое.
the brain is computational.
ведь мозг — своего рода компьютер.
in designing intelligent machinery.
создания искусственного интеллекта.
of those early pioneers,
is not just about accounting
это не только расчёты,
we modeled them after our minds.
похожими на свой мозг.
to understand our own minds better
лучше понять собственный мозг
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