Blaise Agüera y Arcas: How computers are learning to be creative
بلیس آگورا ای آرکاس: رایانهها چطور خلاقیت را یاد میگیرند
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;
که در مورد هوش مصنوعی کار میکند؛
of making computers and devices
برای اینکه رایانهها و دستگاه ها
that brains do.
interested in real brains
in the things that our brains do
to the performance of computers.
از کارایی رایانههاست.
has been perception,
ادراک بوده است،
out there in the world --
for example, that our team makes,
که گروه ما تولید میکنند،
on Google Photos to become searchable,
در گوگل فوتو را قابل جستجو میکنند،
out there into the world.
our work on machine perception
کار ما در ادراک ماشین
with the world of machine creativity
دنیای خلاقیت ماشینی مرتبط شد
had a penetrating insight
between perception and creativity.
ادراک و خلاقیت داشت.
has a statue inside of it,
is to discover it."
Michelangelo was getting at
is an act of imagination
and perceiving and imagining,
بر عهده دارد،
with a brief bit of history
the heart or the intestines,
قلب یا رودهها،
about a brain by just looking at it,
به آن نمیفهمی،
که به مغز توجه کردند
of this thing all kinds of fanciful names,
به ساختارهای سطحی آن دادند،
doesn't tell us very much
اتفاق میافتد نمی دهد.
developed some kind of insight
یک بررسی از آنچه
Santiago Ramón y Cajal,
سانتیاگو رامون ای کاخال بود،
خاصی استفاده کرد
or render in very high contrast
their morphologies.
that he made of neurons
که او از نورونها ساخته
of different sorts of cells,
را میبینید،
was quite new at this point.
در این زمان بسیار جدید بود.
very, very long distances --
تا فاصلههای بسیار زیاد بروند --
to some people in the 19th century;
were just getting underway.
و سیم کشی تازه شروع شده بود.
of Ramón y Cajal's, like this one,
توسط رامون ای کاخال، مثل این،
that Ramón y Cajal started.
شروع کرده را تمام کنیم.
که همکارانمان
of Neuroscience.
تهیه کرده اند.
is about one cubic millimeter in size,
very small piece of it here.
از آن را به شما نشان میدهم.
tiny block of tissue.
of hair is about 100 microns.
much, much smaller
electron microscopy slices,
برشهای میکروسکوپ الکترونی،
in 3D of neurons that look like these.
سه بعدی مثل این را بازسازی کرد.
style as Ramón y Cajal.
کارهای رامون ای کاخال است.
be able to see anything here.
چیزی را ببینیم.
one neuron to another.
ahead of his time,
از زمانهاش کمی جلوتر بود،
over the next few decades.
از الکتریسیته استفاده میکنند،
was advanced enough
فناوری ما به اندازهای پیشرفت کرد
experiments on live neurons
روی نورونهای زنده کنیم
when computers were being invented,
که رایانهها اختراع شدند،
of modeling the brain --
as Alan Turing called it,
که آلن تورینگ نام گذاری کرده بود،
looked at Ramón y Cajal's drawing
به طرح های رامون ای کاخال
imagery that comes from the eye.
از چشم میآید را پردازش میکند.
like a circuit diagram.
in McCulloch and Pitts's circuit diagram
در مدارات مک کلوچ و پیتس وجود دارند
of computational elements
مدارات رایانهای عمل میکند
one to the next in a cascade,
از یکی به دیگری انتقال میدهد،
visual information would need to do.
نیاز داریم صحبت کنیم.
for us to do with our brains.
that for a computer,
just a few years ago.
this task is easy to do.
and the word "bird,"
و کلمه «پرنده» اتفاق میافتد،
connected to each other
inside our visual cortices,
و درون قشر بینایی ما باشد،
to have the capability
on the computer.
that actually looks like.
about as a first layer of neurons,
اولین لایه عصبی باشند،
how it works in the eye --
همانطوری است که چشم کار میکند --
متصل میشوند،
after another layer of neurons,
of different weights.
با وزنهای مختلف متصل میشوند.
of all of those synapses.
مشخص میگردد.
properties of this network.
or a small group of neurons
those three things --
in the neural network,
these synapses in the neural network.
در شبکه عصبیاند.
is just a simple formula,
فقط یک معادله ساده است،
going on there, of course,
of mathematical operations.
that if you have one equation,
اگه یک معادله داشته باشیم،
by knowing the other two things.
دو متغیر دیگه بدست آوریم.
that the picture of a bird is a bird,
تصویر یک پرنده، پرنده است،
and w and x are known.
و X و W معلوم.
you know the pixels.
پیکسلها معلومند.
a relatively straightforward problem.
مسئله سر راسته.
and you're done.
doing exactly that.
که دقیقا همین کار رو میکنه.
on a mobile phone,
روی یک گوشی موبایل،
amazing in its own right,
خودش واقعا عالیه،
billions and trillions of operations
میلیاردها عملیات رو
picture of a bird,
"Yes, it's a bird,"
« آره، این یک پرنده است»،
with a network of this sort.
با شبکهای از این نوع تعیین میکنه.
and the y is the unknown.
و Y مجهول.
difficult part, of course,
do we figure out the w,
of solving for w,
with the simple equation
it's the inverse to multiplication,
معکوس ضربه،
very non-linear operation;
to solve the equation
معادله رو حل کنیم
is fairly straightforward.
a little algebra trick,
to the right-hand side of the equation.
about it as an error.
به عنوان خطا در نظر بگیریم.
for w the right way,
را برای W به شکل صحیح حل کنیم،
to minimize the error,
تا مقدار خطا رو حداقل کنیم،
computers are very good at.
در اون استادند.
sort of play Marco Polo,
successive approximations to w.
but after about a dozen steps,
ولی بعد از نزدیک ده مرحله،
which is close enough.
که به اندازه کافی مناسبه،
a lot of known x's and known y's
و Yهای معلوم رو استفاده میکنیم
through an iterative process.
محاسبه میکنیم.
that we do our own learning.
که ما یادگیری رو انجام میدهیم.
تصویر میبینیم
this is not a bird."
و این پرنده نیست.»
for those neural connections.
x and w fixed to solve for y;
Y مقدارهای X و W رو ثابت نگاه داشتیم؛
Alex Mordvintsev, on our team,
آلکس موردیوینسف، در تیم ما،
with what happens if we try solving for x,
اگر اون رو برای X حل کنیم،
that you've trained on birds,
روی پرنده آموزش دیده رو دارید،
the same error-minimization procedure,
همون شیوه کمینه کردن خطا،
trained to recognize birds,
برای تشخیص پرنده آموزش دیده،
generated entirely by a neural network
یک شبکه عصبی ایجاد شده
rather than solving for y,
by Mike Tyka in our group,
در گروه ما انجام شده،
of William Kentridge's artworks,
ویلیام کنتریج میاندازه،
over the space of different animals,
تغییر Y در فضای حیوانات مختلف است،
to recognize and distinguish
morph from one animal to another.
حیوانی به دیگری تبدیل میشه رو میبینید.
have tried reducing
out of the space of all things
همه چیزهایی بسازند
over that entire surface,
you make a kind of map --
یه نوع نقشه تولید میکنی --
the network knows how to recognize.
که شبکه میتونه تشخیص بده.
"armadillo" is right in that spot.
«گورکن» درست اونجاست.
of networks as well.
با انواع دیگر شبکه انجام بدی.
to recognize faces,
چهره طراحی شده،
in a y that says, "me,"
که بیانگر «من» هست،
psychedelic picture of me
شبیه سبک کوبیسم، سور رئال
multiple points of view at once
تولید شده
to get rid of the ambiguity
or another pose,
another kind of lighting.
this sort of reconstruction,
of different points of view,
his own face as a guide image
صورت خودش به عنوان راهنما
to reconstruct my own face.
برای ایجاد صورت من استفاده کرده.
that optimization process.
انجام شود.
more like a coherent face,
شبیه به صورتی منسجم بدست آوریم ،
with a blank canvas
that is itself already some other image.
که خودش تصویر دیگری است.
that is designed to categorize
man-made structures, animals ...
ساختههای انسان، حیوانات ...
with just a picture of clouds,
what it sees in the clouds.
که در ابرها میبیند را پیدا میکند.
you spend looking at this,
will see in the clouds.
to hallucinate into this,
برای خبال پردازی در آن استفاده کرد،
zooms hallucinates, zooms.
بزرک نمایی.
of the network, I suppose,
در یک وضعیت گیجی شبکه قرار میگیری ،
is eating its own tail.
What do I think I see next?"
فکر میکنم بعدی چیه؟» خواهد بود
called "Higher Education" --
در سیاتل این را نشان دادم --
marijuana was legalized.
is not constrained.
این فناوری هنوز محدود نشده.
because they're really fun to look at.
آن را نشان دادم چون دیدنشان لذت بخش است.
a camera that takes a picture,
دوربینی عکس می گیرد،
writes a poem using neural networks,
در موردش با شبکه عصبی شعری میگوید،
has been trained
از اشعار قرن بیستم آموزش دیده.
are very intimately connected.
ذاتا بسیار به هم نزدیکند.
things in the world,
Michelangelo really did see
any being, any alien
هر موجود فرازمینی
perceptual acts of that sort
فعالیتی ادراکی داشته باشد
machinery that's used in both cases.
هر دو کار را انجام میدهد.
and creativity are by no means
ادراک و خلاقیت به هیچ وجه
that can do exactly these sorts of things.
که دقیقا اینگونه کارها را انجام میدهند.
the brain is computational.
چون مغز هم محاسباتی است.
in designing intelligent machinery.
طراحی دستگاههای هوشمند کرده است.
دستگاهها را هوشمند کنیم انجام شده.
of those early pioneers,
این پیشگامان،
is not just about accounting
we modeled them after our minds.
بر مبنای ذهنمان ساخته ایم.
to understand our own minds better
تا ذهنمان را بهتربشناسیم
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