WILLIAM BRANGHAM: The dream of artificial intelligence, that is, machines that think like humans, is starting to become a reality.
Now, this development triggers a whole host of concerns about our jobs, about what's real and what isn't, and especially about whether we will be able to control a technology that suddenly has a very smart mind of its own.
But it's not all as grim as it seems.
As science correspondent Miles O'Brien reports, A.I.
may create new tools to address some of the most complex problems, and at the top of that list is the climate emergency.
MILES O'BRIEN: As the world grapples with the urgent threat of climate change, scientists and policymakers are turning to an unlikely ally, artificial intelligence.
From predicting extreme weather events to optimizing energy use, A.I.
is emerging as a powerful tool in the fight against global warming.
MAN: We fly over a scene of interest as detected by the camera connected to the A.I.
chip.
MILES O'BRIEN: In Orlando, Florida, a small weather forecasting company called MyRadar is working to combine some tiny satellites with artificial intelligence to detect fires long before they race out of control.
ANDY GREEN, Founder and CEO, MyRadar: Yes, it's not -- I would worry how long the boot-up time is to be.
MILES O'BRIEN: That's founder and CEO Andy Green.
ANDY GREEN: We're employing A.I.
to do something that a human can't.
We want it to look at the planet below on a 24-by-7 basis to look for disasters in the making.
MILES O'BRIEN: A confluence of technologies, A.I., coupled with the relentless miniaturization of sensors and the reduced cost of reaching low-Earth orbit, put this company on a trajectory to launch its own satellite constellation designed to spot fires early.
Sarvesh Garimella is MyRadar's chief scientist and CTO.
SARVESH GARIMELLA, CTO, MyRadar: Getting that early information and getting the fire identified and tracked is the challenge that we're trying to solve from space.
MILES O'BRIEN: The existing fleet of satellites flown by the National Oceanic and Atmospheric Administration and NASA are not ideally suited for real-time alerting of a budding fire.
Either the image resolution is not granular enough or the satellite doesn't pass over a particular site frequently enough.
SARVESH GARIMELLA: We're filling in the gaps by trying to launch low-Earth orbit satellites that will bring the revisit time down from 12 over 24 hours to sub-hourly.
MILES O'BRIEN: NOAA is providing $800,000 in funding.
The company is aiming to launch the first of its armada in October of 2024.
ANDY GREEN: The intent is to build out a full constellation of roughly 150 to 200 satellites or more, and that will give us the coverage that we need and the timeliness of the data that we want to get.
MILES O'BRIEN: The idea is to give users of the MyRadar app, including first responders, real-time warnings of fires.
SARVESH GARIMELLA: This one is the near-I.R.
hyperspectral camera.
MILES O'BRIEN: They will be equipped with high-resolution visible light and near-infrared hyperspectral cameras, as well as a thermal imager.
You can put that much into what is essentially a Rubik's Cube.
SARVESH GARIMELLA: Indeed, yes, so 10 centimeters on a side.
MILES O'BRIEN: That's kind of extraordinary, isn't it?
SARVESH GARIMELLA: It is.
It's really exciting.
MILES O'BRIEN: They call it HORIS, the Hyperspectral Orbital Remote Imaging Spectrometer.
In addition to the sensors, these cubesats will fly with artificial intelligence on board.
It's a way to solve a communications bottleneck.
SARVESH GARIMELLA: The onboard A.I.
piece of it means that, once we do detect a fire, we can process that information on board the satellite and send an alert directly to the ground without having to downlink the entire data set for somebody on the ground to process themselves.
A.I.
on board makes the mission possible.
MILES O'BRIEN: So what makes the A.I.
able to see fires so early?
It's a perfect use case for a so-called convolutional neural network.
Here's how it gets smart enough to spot specific things.
Take a dog, for example.
It combs through a picture with many virtual magnifying glasses.
Each one is looking for a specific kind of puzzle piece, like an edge, a shape or a texture.
Then it makes simplified versions, repeating the process on larger and larger sections.
Eventually, the puzzle can be assembled and it's time to make a guess.
Is it a cat, a dog, a tree?
Sometimes, the guess is right, but, sometimes, it's wrong.
But it learns from mistakes.
Labeled images are sent back to correct the previous operation, so the next time it plays the guessing game it will be even better.
ANDY GREEN: Ultimately, you get to a point where you can show a picture of a cat and say, is this a cat?
And then it will reply back to you and give you a yes-or-no answer with some degree of confidence.
It's essentially the same process for wildfires.
We will train the model that we use based on some existing data, and that provides it a good general understanding of what a wildfire might look like or what a smoky environment might look like.
And so, ultimately, in the end, we have a simple numerical model that says, this looks like a wildfire could be in place.
Let's alert somebody on the ground and take a look at it and potentially prevent it from spreading further.
MILES O'BRIEN: Artificial intelligence is not just being deployed for climate adaptation.
It is also a potent tool for mitigation.
At the Lawrence Berkeley National Laboratory in California, they are applying A.I.
to the urgent hunt for green energy sources.
Is there anything like this in the world?
GERBRAND CEDER, Lawrence Berkeley National Laboratory: Not in inorganic chemistry.
MILES O'BRIEN: Material scientist Gerd Ceder showed me the place he calls the A-lab.
GERBRAND CEDER: We deliberately did not define what the A stands for.
MILES O'BRIEN: So there is no B-Lab?
GERBRAND CEDER: Could stand for automated, autonomous or A.I.-driven.
MILES O'BRIEN: A.I.-driven robotic lab technicians are at work here around the clock testing recipes for compounds that might make better batteries to enable the transition to renewable energy.
Historically, it's been slow going, tedious trial and error.
GERBRAND CEDER: Somebody comes up with an idea, goes and tries it in the lab, iterates on that many, many, many times, and that's why it takes so long.
The average time to market is somewhere between 18 and 20 years.
MILES O'BRIEN: So how to speed things up?
GERBRAND CEDER: And what's the different branches?
MILES O'BRIEN: The solutions they seek are buried in millions of scientific papers.
GERBRAND CEDER: OK, so we should have results tomorrow or so.
MILES O'BRIEN: The lab has developed machine learning algorithms that sift through nearly all of the scientific literature on material science.
The A.I.
can see correlations and anomalies humans might not.
This allows it to predict the properties of vast numbers of hypothetical compounds, reducing the need for trial and error in the lab and saving time and resources.
GERBRAND CEDER: A person can never know what's been done in five million research papers, but that's the beauty of mathematics and of computing algorithms.
They essentially hold all knowledge in memory.
MILES O'BRIEN: To more efficiently test hypothetical compounds, they had built this robotic lab to mix and test the suggestions 24/7.
Ceder says what used to take months now happens in a matter of days.
GERBRAND CEDER: We really can innovate materials at a much, much faster scale.
We can't come up with solutions in 40 years, right?
We just have to get on with it.
MILES O'BRIEN: For many people, artificial intelligence is this.
ARNOLD SCHWARZENEGGER, Actor: Come with me if you want to live.
MILES O'BRIEN: The Terminator, an existential threat.
But it might very well be an indispensable tool to confront the complex problems that threaten us most.
For the "PBS NewsHour," I'm Miles O'Brien in Berkeley, California.
WILLIAM BRANGHAM: Miles and his team have been working on this subject for nearly a year now, and the result is a one-hour film called "A.I.
Revolution."
And it premieres tonight on "NOVA" at 9:00 p.m. Eastern here on PBS and available for streaming online.
