[whale whistles]

[whale chirps]

[whale bellows]

[whale whistles]

Clark: Everything in the ocean is producing some form of sound.

I'm Christopher Clark,

I'm a bioacoustic scientist,

and I listen to the songs of life

around this planet.

[whale bellows]

Clark: You listen to this expanse of the sound in the ocean.

It just becomes a magnificent symphony of voices,

and you know what?

Most of the voices that we're listening to today

we do not know who's making those sounds.

[curious music]

Allen: My name's Ann Allen.

I'm a research oceanographer

for NOAA Fisheries.

We use acoustics or sound in order to help us monitor

the whales and dolphins in the Pacific Islands.

[curious music continues]

[whale bellows]

[whale cries]

Allen: Pretty early on, I realized

that a lot of the methods that have been developed so far

were not going to work for our data.

Somebody would literally sit here and scan through a few hours at a time

marking each of these start and ends of the humpback song.

[whale bellows]

[whale chirps]

Harvey: Hi, my name's Matt Harvey,

I'm a software engineer at Google,

and I work on machine learning models for audio analysis

applied to bio acoustics.

Allen: Some species like fin whales make very simple sounds,

and our detectors are actually pretty good at those,

but then species like humpback whales make very complex sound types

that are changing all the time,

so having a detector that works for those sounds

is very, very difficult.

[whale cries softly]

[whale groans]

Allen: The idea with machine learning is, you train a machine learning model,

which is actually teaching a computer to recognize sounds

rather than teaching it a step-by-step process.

And this is humpback whale. This is humpback whale.

This is not humpback whale. Ignore this.

Harvey: We learned to predict the presence of humpback whale

with very high accuracy.

Allen: So it's doing what it's supposed to do.

It's doing what it's supposed to do, yeah.

Allen: Whoo!

There's not a lot known about our whale populations out here,

so there's so much that we could figure out

just from knowing who's where and when.

[whale bellows]

Allen: We've collected all this data.

We don't want to just hoard it.

We want to put it out there so that other people can use it and look at it.

Jongejan: When I heard the first audios of underwater recording,

I was amazed by that world that we typically don't hear

and I want to share that with more people.

Hi, my name is Jonas,

and I'm a creative technologist at Google.

[gentle music]

Jongejan: So the data you see is all visualized here as a spectrogram.

When you zoom all the way in, you would see a few minutes of audio,

and you can see the patterns of the different whale calls.

As you zoom out, you would be able to see several months of audio

at the same time on the screen.

And at the bottom, you see a long bar that is a heat map,

so using the AI to guide you where to look for whales.

Clark: It's absolutely critical that we learn to share.

What you're doing is, you're actually allowing a vista

into a world that's been totally hidden from us.

[whale bellows]

[whale shrieks]

[whale groans]

[whale chatters]

Allen: There's sounds in data sets that we don't know what they are.

Jongejan: We've seen that releasing a large data set

just creates opportunities that we have never thought about.

People might take this project

and make discoveries

that no one else has done before.

[whale squeaking]

Clark: You're opening up the opportunity to explore it to everybody.

Now you're not going to have 10 people or even 100 people.

You're going to have 10 million people.

[gentle music]

Clark: That's the scale we need to understand life in the ocean.

[whale chirps]

[whale bellows faintly]

[whale chirps]

[whale squeaking]