Blaise Agüera y Arcas: How computers are learning to be creative
Blaise Agüera y Arcas: Cum învaţă calculatoarele să fie creative
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;
se ocupa de inteligenţa artificială;
of making computers and devices
a creării de computere şi dispozitive
that brains do.
din lucrurile făcute de creier.
interested in real brains
de creierele reale şi de neurostiinţă,
in the things that our brains do
pe care creierele noastre le fac
to the performance of computers.
performanţelor computerelor.
has been perception,
a fost percepţia:
out there in the world --
lucruri din lumea reală
for example, that our team makes,
de percepţie artificială creaţi de noi,
on Google Photos to become searchable,
din Google Photos să poată fi căutate,
out there into the world.
în ceva real, concret.
our work on machine perception
în domeniul percepţiei artificiale
with the world of machine creativity
şi cu lumea creativităţii artificiale
had a penetrating insight
a avut o intuiţie profundă
between perception and creativity.
dintre percepţie şi creativitate.
has a statue inside of it,
are o statuie în interiorul său,
is to discover it."
e să o descopere”.
Michelangelo was getting at
is an act of imagination
e un act de imaginaţie
and perceiving and imagining,
şi imaginaţie e, desigur, creierul.
with a brief bit of history
the heart or the intestines,
about a brain by just looking at it,
despre un creier doar uitându-te la el,
care se uitau la creierul uman
of this thing all kinds of fanciful names,
ale acestuia tot felul de nume fanteziste,
doesn't tell us very much
nu ne spun mare lucru
developed some kind of insight
cred că a înţeles oarecum
Santiago Ramón y Cajal,
din secolul 19, Santiago Ramón y Cajal,
şi coloraţii speciale
or render in very high contrast
să redea în contrast foarte înalt
their morphologies.
să le înţeleagă morfologia.
that he made of neurons
făcute de el în secolul 19.
of different sorts of cells,
de diverse tipuri de celule,
was quite new at this point.
destul de nouă la acel moment.
very, very long distances --
la distanţe foarte mari...
to some people in the 19th century;
pentru unii din secolul 19;
were just getting underway.
şi electricităţii tocmai începea.
of Ramón y Cajal's, like this one,
Ramón y Cajal, precum acesta,
that Ramón y Cajal started.
începută de Ramón y Cajal.
de la colaboratorii noştri
of Neuroscience.
de Neuroştiintă Max Planck.
is about one cubic millimeter in size,
are cam un milimetru cub,
very small piece of it here.
doar o foarte mică parte din ea.
are aproape un micron.
sunt mitocondrii,
tiny block of tissue.
of hair is about 100 microns.
e de aproximativ 100 de microni.
much, much smaller
electron microscopy slices,
de microscopie electronică,
in 3D of neurons that look like these.
ale neuronilor care arată aşa.
style as Ramón y Cajal.
ale lui Ramón y Cajal.
be able to see anything here.
distinge nimic aici.
one neuron to another.
neuron de celălalt.
ahead of his time,
cu un pas înaintea epocii sale,
over the next few decades.
folosesc electricitate
era destul de avansată
was advanced enough
experiments on live neurons
electrice pe neuroni vii
cum funcţionează.
when computers were being invented,
erau inventate computerele,
of modeling the brain --
modelării creierului
as Alan Turing called it,
cum a numit-o Alan Turing,
looked at Ramón y Cajal's drawing
s-au uitat la desenul lui Ramón y Cajal
– pe care îl vedeţi aici.
imagery that comes from the eye.
imaginile venite de la ochi.
like a circuit diagram.
ca diagrama unui circuit electric.
in McCulloch and Pitts's circuit diagram
lui McCulloch şi Pitts
of computational elements
ca o serie de elemente computaţionale
one to the next in a cascade,
unul către altul în cascadă,
visual information would need to do.
al procesării informaţiei vizuale.
for us to do with our brains.
pentru creierul nostru.
that for a computer,
că pentru un computer,
just a few years ago.
this task is easy to do.
e uşor de îndeplinit.
and the word "bird,"
şi cuvântul „pasăre”,
connected to each other
într-o reţea neuronală,
inside our visual cortices,
poate fi biologică, în cortexul vizual,
capacitatea să modelăm
to have the capability
on the computer.
pe calculator.
that actually looks like.
about as a first layer of neurons,
ca primul strat de neuroni,
how it works in the eye --
mai departe
after another layer of neurons,
of different weights.
de diferite dimensiuni.
of all of those synapses.
tuturor sinapselor.
properties of this network.
computaţionale ale acestei reţele.
or a small group of neurons
sau un grup mic de neuroni
those three things --
in the neural network,
din reţeaua neuronală,
these synapses in the neural network.
sinapselor din reţeaua neuronală.
is just a simple formula,
going on there, of course,
of mathematical operations.
de operaţii matematice,
that if you have one equation,
by knowing the other two things.
dacă le ştii pe celelalte două.
that the picture of a bird is a bird,
că imaginea unei păsări e o pasăre
and w and x are known.
w şi x sunt cunoscute.
you know the pixels.
cunoaştem pixelii.
a relatively straightforward problem.
e o problemă relativ simplă.
and you're done.
neuronală artificială
doing exactly that.
on a mobile phone,
pe un telefon mobil,
amazing in its own right,
billions and trillions of operations
miliarde şi trilioane de operaţii/secundă.
picture of a bird,
ale unei păsări,
"Yes, it's a bird,"
with a network of this sort.
folosind o reţea de acest tip.
and the y is the unknown.
iar y e necunoscut.
difficult part, of course,
do we figure out the w,
of solving for w,
al aflării lui w,
with the simple equation
it's the inverse to multiplication,
e inversul înmulţirii,
very non-linear operation;
foarte non-lineară,
to solve the equation
de a rezolva ecuaţia
is fairly straightforward.
a little algebra trick,
to the right-hand side of the equation.
în partea dreapta a ecuaţiei.
about it as an error.
ca pe o eroare.
for w the right way,
în modul corect,
to minimize the error,
pentru a reduce eroarea.
computers are very good at.
la care computerele excelează.
sort of play Marco Polo,
ar putea să o ia pe bâjbâite,
successive approximations to w.
obţine aproximări succesive ale lui w.
but after about a dozen steps,
dar cam după o duzină de paşi,
which is close enough.
ceea ce e destul de aproape.
a lot of known x's and known y's
de x şi y cunoscuţi
through an iterative process.
printr-un proces iterativ.
that we do our own learning.
this is not a bird."
asta nu e o pasăre”.
for those neural connections.
conexiunile neuronale.
x and w fixed to solve for y;
pentru a-l afla pe y;
prin multe încercări de învăţare.
Alex Mordvintsev, on our team,
din echipa noastră,
with what happens if we try solving for x,
dacă încercăm să descoperim x,
that you've trained on birds,
care a fost pregătită pe păsări,
the same error-minimization procedure,
de minimizare a erorii,
trained to recognize birds,
să recunoască păsări,
generated entirely by a neural network
generată integral de o reţea neuronală
rather than solving for y,
în loc să-l afle pe y
by Mike Tyka in our group,
în grupul nostru,
of William Kentridge's artworks,
ale lui William Kentridge,
over the space of different animals,
în funcţie de diferite animale,
to recognize and distinguish
şi să distingă diferite animale,
morph from one animal to another.
de tip Escher, de la un animal la altul.
have tried reducing
au încercat să-l reducă pe y
out of the space of all things
din spaţiul tuturor lucrurilor
over that entire surface,
pe întreaga suprafaţă,
you make a kind of map --
faci un fel de hartă,
the network knows how to recognize.
pe care reţeaua le recunoaşte.
"armadillo" is right in that spot.
Tatuul (armadillo) e în punctul ăla.
of networks as well.
şi cu alte tipuri de reţele.
to recognize faces,
in a y that says, "me,"
psychedelic picture of me
suprarealista, psihedelică
multiple points of view at once
puncte de vedere simultan
to get rid of the ambiguity
ca să elimine ambiguitatea
or another pose,
another kind of lighting.
this sort of reconstruction,
of different points of view,
a mai multor puncte de vedere
his own face as a guide image
chipul său ca imagine de control
to reconstruct my own face.
pentru reconstrucţia feţei mele.
that optimization process.
acestui proces de optimizare,
more like a coherent face,
ceva ce aduce mai mult cu o faţă,
with a blank canvas
that is itself already some other image.
e el însuşi o altă imagine.
that is designed to categorize
man-made structures, animals ...
structuri create de om, animale...
with just a picture of clouds,
what it sees in the clouds.
ce vede în nori.
you spend looking at this,
will see in the clouds.
veţi vedea şi voi în nori.
to hallucinate into this,
pentru a halucina astfel,
zooms hallucinates, zooms.
măreşte, halucinează, măreşte.
of the network, I suppose,
a reţelei, presupun,
is eating its own tail.
What do I think I see next?"
Ce cred că văd acum?”
asta în public
called "Higher Education" --
intitulat „Educaţia Superioară”.
marijuana was legalized.
după legalizarea marijuanei.
is not constrained.
nu e restricţionată.
because they're really fun to look at.
deoarece sunt distractive.
artistul Ross Godwin,
a camera that takes a picture,
care face o fotografie
writes a poem using neural networks,
o poezie folosind reţelele neuronale,
has been trained
a fost antrenată
de poezii al secolului 20.
după părerea mea.
are very intimately connected.
sunt puternic conectate.
pregătite să distingă
things in the world,
diferite lucruri din lume,
ca să genereze.
Michelangelo really did see
blocului de piatră,
any being, any alien
orice extraterestru
perceptual acts of that sort
de acte perceptive de acel gen
machinery that's used in both cases.
e folosit în ambele cazuri.
and creativity are by no means
creativitatea nu sunt exclusiv umane.
that can do exactly these sorts of things.
care fac exact genul ăsta de lucruri.
the brain is computational.
creierul e computaţional.
in designing intelligent machinery.
de creare a inteligenţei artificiale.
of those early pioneers,
primilor pioneri,
is not just about accounting
nu mai însemnă doar contabilitate,
we modeled them after our minds.
după minţile noastre.
to understand our own minds better
mai bine propriile minţi
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