Name: データを”絵の具”にする男。機械学習が創る人類の記憶 | Obsessed | WIRED Japan
Uploaded: 2025-01-15T08:40:24.000Z
Duration: 10 min 33 s
Description: 【看影片學英語】數萬部 YouTube 影片，搭配英漢字典即點即查，輕鬆掌握單字發音與用法，長久累積看電影不必再看字幕。

基礎被動語態

It's easy to get lost in Rafiq Anadol's surreal installations, but dig a little deeper and you realize his work is made up of millions of tiny pieces, and that every single little point represents a piece of data that he's fed through a neural network to show us his vision of our future.

I was always like trying to like speculate this idea of can data become a pigment.

情態副詞

At the end the data is a truly numbers that has no kind of any skin or skeleton, but what I'm trying to as an artist is find the algorithm that can narrate the moment of data, kind of like make that invisible moment visible.

Enormous data sets inspire most of Anadol's art, and he uses machine intelligence and algorithms to create visualizations, what he calls data sculptures, like this piece called Machine Hallucinations.

He started by finding 113 million images of New York online.

And let the machine learn from this entire corpus of data.

And then we also find a way to erase the humans from this data and only focus on the buildings, nature, and environments that is a collective memory of New York.

Once all of the images of people were removed, Anadol was left with 10 million pictures of New York.

He fed them through a machine learning algorithm that generates visual associations as it learns.

For example, when it sees multiple photos of the Statue of Liberty taken from slightly different angles, the algorithm interpolates information to help it create a moving, shifting image that represents the entire life cycle of the structure.

Machine looks at this information like a human being, but it's kind of more like a collective memory than a personal memory.

Because a building in New York can be explored by thousands of perspectives from multiple angles from a different time of the year.

It's more like an honest memory for a machine because it feels more totalitarian and feels everything and everyone than just one person.

Anadol covered the walls and the floor of a boiler room in a building in lower Manhattan with this machine-interpreted dynamic landscape.

If you ask him, he'll tell you that the machine is dreaming.

When a machine learns from outputs and memories like this, it can create an alternative reality.

It looks at the patterns of the trees, the buildings, the nature, the people, every single thing hidden inside this image corpus.

Seeing a machine giving a context of data and creating hallucinative outputs was something like really inspiring.

Anadol created this piece to celebrate the centennial of the LA Philharmonic, using half a million images, thousands of audio recordings, and hundreds of videos from the orchestra's  He fed all that into a series of algorithms that turned it into these extraordinary projected images.

By using machine learning algorithms, the entire archive of LA Philharmonic becomes a shape of three-dimensional outputs.

We were able to see all these data points becoming together or disappearing and making kind of a new sculpture.

And the building itself, Frank Gehry's Walt Disney Concert Hall, became part of the sculpture, too.

I was always looking for inspiring buildings, and Frank Gehry became my hero.

And that night, I was really struck by an idea, like, what will happen if this building can remember?

So he projected the machine-interpreted archives onto a smaller model of the Walt Disney Concert Hall.

Once he settled on something he liked, he projected it onto the bigger, real-life building.

What we are doing is not a typical painting, not a typical sculpture, but more like an experience from a near future.

And when they come together with sound, data, machine intelligence, and light, and architecture, they have a new meaning, a new kind of symbiotic connections.

I think I was really hooked up by the idea of a building can become an interface.

It can be a boring advertising, but it can be a beautiful interface.

The ideation of a human instinct become a part of a machine, it really strikes me, like,  Anadol and his team used 42 large-scale projectors with an astonishing 50K video resolution to create a dramatic display in downtown LA every night for 10 days, a nod to one of his favorite movies.

I was eight years old when I watched the movie Blade Runner.

In downtown Los Angeles, a building suddenly becomes alive.

And it has this cognitive capacity of remembering, and it has a capacity of dreaming.

So this was all, like, very high-level science fiction narratives.

And the futuristic dystopian film's exploration of more complex themes echo Anadol's interest in the relationship between humans and technology.

When Deckard and Rachel, like, contemplates each other and has a dialogue where Deckard defines Rachel as a replicant, it was a very interesting moment where a human defines another machine that is kind of defining what is real or what is not.

That kind of human consciousness and machine consciousness were having its dialogue moment.

You remember the spider that lived in the bush outside your window?

Those aren't your memories, they're somebody else's.

I think in humanity, we have very certain findings that transform who we are.

When we found the fire, we cook with it, we create communities.

With the same technology, we kill each other and destroy.

And clearly, AI is one of the discoveries in humanity that has a potential to make communities or destroy each other.

Anadol finds data for his machine intelligence collaborations everywhere.

And when he learned that wind data is continuously being collected at the Boston airport, he knew he'd found a goldmine.

He turned it all into this project called Winds of Boston.

I was always inspired by nature as an inspiration, from the wind itself, the water, the nature, like a core natural aspect of motion theory in life in general.

And I thought it can be an incredible opportunity to use wind as a data and as a pigment.

So we first located our data source from the airport, Logan Airport.

The wind is extremely important for the flight networks.

So we took a one-year-long wind data of Boston.

And this data consists of the gust, speed, and direction, and the temperature of the weather.

Anadol fed all this wind data into a series of algorithms.

Then he built custom 13-foot-tall LED screens to display the visualized data.

And I thought that algorithms can be an incredible way of visualizing this invisible pattern of wind and transform them into a poetic-like motion.

Because it's very inspiring that a machine can find something interesting that I didn't even think about it.

And I found that the true collaboration starts happening there.

And sometimes machines give opportunities that I didn't think about it.

And I do believe that it's really something that is inspiring for humanity that will most likely bring some new kind of imagination methods that is not discovered yet.

The inspiration for this project, called Melting Memories, came from a personal place.

I went to Istanbul, my hometown, and my uncle just couldn't remember where I'm coming from.

And apparently it was his first stage of Alzheimer's.

When this disease happens, we literally melt our memories.

But I also was very curious about scientifically, honestly, what memory means.

What is a cognitive representation of a memory?

And the images on these 20-foot tall LED screens represent the real data behind that process.

And these moments are really trying to give a sense of a tangible feeling of a memory.

I know that we are not there as technology that truly transcripts memory, but it at least gives a glimpse of an abstract language of a memory.

That's one of those feelings that is hidden in those algorithmic exploration of data.

Anadol partnered with scientists at a neuroscience lab.

They asked subjects to concentrate on childhood memories and recorded all their brain pulses using an EEG.

The data itself is literally pulsing in four milliseconds of human brain activity in the two locations of brain, from hippocampus to frontal left lobe.

And the algorithm symbolizes the moment of remembering, and it is how poetic moment of this firing location in two locations of the brain.

Anadol and his team created custom software that transformed all of that brain data into an artistic interpretation of the neurons firing.

Another example of how he uses technology to describe life.

I never stopped thinking about data as a mean of material.

Sometimes it can be wind data, it can be a Wi-Fi, Bluetooth signal.

It can be a machine intelligence, machine decisions.

It's honestly anything that machine needs to understand life can become a material for this imagination.

One problem we have to solve today is a visual discontinuity problem.

Anadol has built a team with expertise in different fields to help him fulfill his vision.

AI experts, computer scientists, architects, and designers.

And they're working on a new workflow for a data-driven public art project set for Portland.

For this one, Anadol is letting hundreds of thousands of images of Portland inform both the projections and the structure they'll be projected on.

The team is using a 3D printer to build the structure, one panel at a time.

It will ultimately be a 21-foot-tall sculpture.

I found that we were stuck in this just virtual world.