The Wall Street Journal

Police Unlock AI's Potential to Monitor, Surveil and Solve Crimes


Law enforcement agencies like the New Orleans Police Department are adopting artificial-intelligence based systems to analyze surveillance footage. WSJ's Jason Bellini gets a demonstration of the tracking technology and hears why some think it's a game changer, while for others it's raising concerns around privacy and potential bias. Photo: Drew Evans/The Wall Street Journal

Tampa Bay Buccaneers star Rob Gronkowski sits down with Randy Moss to discuss why he ultimately decided to return to the NFL, how it's been playing with Tom Brady again and shares how he has had to adjust playing on a new team.

During the pandemic many older members of society have been separated from and unable to meet children and younger members of the family.

97-year-old Margot Lawson has met her great granddaughter for the first time, viewing her through an acrylic screen.
With less than two days to go until President Trump's term ends, a senior administration official tells CBS News Mr. Trump plans to issue 50 to 100 pardons and commutations Tuesday. CBS News chief White House correspondent Nancy Cordes, CBSN political reporter Caitlin Huey-Burns, and Washington Post political reporter Eugene Scott spoke to CBSN's "Red & Blue" host Elaine Quijano about the tense mood in Washington ahead of Inaguration Day.
Krystal Ball and Saagar Enjeti discuss leaked documents from Amazon's Global Security Operations Center that show the company's reliance on Pinkerton operatives to spy on warehouse workers.

… show captions ↓
- [Narrator] Facial recognition has become
a hot button issue in the U.S.
The controversial practice
was recently banned in San Francisco.
But few people realize that many police departments
have adopted artificial intelligence,
using it to analyze video surveillance footage,
which can track objects and people.
- I've taken 8,213 objects,
and I've reduced it to 80 objects.
- [Narrator] Other new technologies, like sensors,
drones equipped with augmented reality capabilities,
and real-time crime centers are also making their way
into police departments in the U.S.
- All that technology, used for the right purposes
and the right reasons, I think is good.
- So while, no, I don't want to be mugged,
I want to be in a community that's safe,
I don't think expanding surveillance is the way to do that.
- There is just a lot of technology being acquired
that's collecting information on everyone,
regardless of whether you're involved in a crime or not.
And I don't want to end up in a version of America
where people are afraid to do things in life
because of being watched all the time.
- [Narrator] To understand how police departments
use these new technologies which makes law enforcement
much more powerful, we traveled to New Orleans.
For the past year and a half,
the city's Office of Homeland Security
has been operating a real-time crime center.
The city's installed over 400 cameras in hotspots,
from the French Quarter to outer-lying districts.
It also pulls in camera feeds from over 150 businesses
and private home owners.
The system screens all incoming calls for service,
including 911 calls, which automatically
activate the network of cameras.
- All of these cameras would spin up automatically
in the software when any incident is created
within a tenth of a mile of these three blocks.
- So your eyes can be there before an officer's eyes.
- Sure. All of our cameras are pan, tilt zoom cameras,
30 times optical zoom.
- [Jason] And you can very easily rewind back
and also see it happening.
- Sure. If it's obviously associated with a criminal matter,
we'd go through historical footage
and provide that information to the officers.
- Some police officers in New Orleans
have quickly embraced this access
to recorded surveillance cameras.
(lively jazz music)
This street's pretty well covered by cameras?
- Yes.
Cameras on every intersection.
- Paul Johnson's been with the New Orleans
police force for six years.
He's a night shift detective in the French Quarter.
That camera right there,
it's hiding in plain sight right there.
- Correct, yes.
This particular camera actually has assisted me
on one of my latest cases.
Probably like a couple months ago
we had a shooting at this very intersection
about 4:30 in the morning.
The gentleman that got shot claimed to be the victim.
We watched the video surveillance, went back,
and it showed that actually he was actually
shooting at the young man first.
So it kind of helped us exonerate
the guy we were looking for.
- And if you didn't have the camera,
do you think you would have known the truth?
- No, absolutely not.
(crowd cheering)
- We talked to Paul Noel, the second in command
of the New Orleans police.
There are critics out there and people
who aren't comfortable with the idea
of the police having cameras.
- Absolutely.
The more information that we have out there,
we need to make sure we're using that information
for the appropriate purposes.
That's one thing where we absolutely stress
here in New Orleans, is that we have accountability measures
in place with all the information that we have,
that the information is not gonna be abused
by our investigators.
- [Jason] One of the critics questioning
the surveillance cameras in New Orleans
is community activist Dee Dee Green.
She says that some people in her group, Stop Watching Nola,
are afraid that they're being monitored,
since a camera was placed right next
to their community garden.
- Knowing that the city is observing you and monitoring you
may make people uneasy and uncomfortable
about having a public conversation
about the politics of the city.
- In your neighborhood is there a consensus
that these cameras shouldn't be here?
- No, I don't think so at all.
I attend a lot of neighborhood association meetings.
People will say we need a camera on this corner,
because this is a crime corner,
or this is a problem corner.
It's very easy to sell the narrative
that these are here for your protection.
We know that people of color, black and brown folks,
are disproportionately profiled and targeted
by police, arrested, incarcerated.
So while, no, I don't want to be mugged,
I want to be in a community that's safe,
I don't think expanding surveillance
is the way to do that.
- [Jason] What most people like Dee Dee Green don't know
is that not only can the video be watched and played back;
the system's now equipped with machine learning
and can analyze the content of recordings itself
without any human oversight.
The crime center in New Orleans
uses a product called BriefCam.
- In your case, you simply said "red truck".
So if I changed the class of this to pickup
and the color to red, I've taken the 8,213 objects
that went through this space in a matter of two hours,
and I've reduced it to 80 objects,
and I can finish watching the totality of it
in a minute and 45 seconds.
- Superimposing 'em on top of one another.
- [George] It's bringing in that actual object
onto a fixed still frame
and showing its total path throughout the frame.
- [Jason] More and more law enforcement
and public safety agencies across the U.S.
are buying these kinds of technologies.
Motorola Solutions is one of the largest
vendors in the space.
In Plantation, Florida, the company's showing us
some of its newest high-tech gadgets and technologies,
which the company only gives
to potential law enforcement customers.
The demo also includes a drone-mounted camera.
Motorola Solutions sells a system
that augments drone footage with real-time information,
like labels identifying each officer on the scene.
Daniel Tialdi from Motorola Solutions
is playing a suspected criminal, and I'm a police officer
at a private testing ground that's closed to the public.
- [Designer] So this is you walking up on the scene,
and we're tracking you from the drone feed.
- [Jason] Tialdi attacks and stabs me.
When this smart vest is pierced,
the embedded sensors automatically notify the command posts.
- So they can send reinforcements.
- [Jason] It also switches on the body-worn camera.
- Because your body camera came on,
we actually have a description of who stabbed you.
- [Jason] An analyst from Motorola shows us how,
using the suspect's description,
he can electronically hunt him down.
- We know that it was a man.
We know that he was wearing a red shirt.
We know generally what time it was.
As we start seeing all these people in red shirts,
we can start looking for the actual individual
that we're looking for.
- [Jason] Motorola Solutions calls this
descriptive artificial intelligence.
The machine can pull from the pixels
a description of a truck, a car, or a person.
- You can see our results get better and better.
As we get more of them, we see more and more.
It also tries to isolate and pull out
a clean shot of the face, the face tile as it's called.
- The face tile can be used to refine the search
by looking for that same face
on other cameras connected to the system.
If you have a long enough period of time,
you could begin to see patterns of behavior as well.
- Most definitely, most definitely.
- [Jason] Once it knows what to look for,
the system can also search backwards in time
to track a particular individual's prior locations.
This virtual time machine, experts say, is a game changer.
It could, in many cases, eliminate the need
for officers to track a suspect
in person for days on end.
Some of these new video analytics tools
can detect weapons or unusual behavior.
"Direction violated".
- Yeah, this is your typical one,
so if you're at the airport,
you're coming out of these TSA areas.
If anybody's walking the direction
of the arrows, they're fine.
If they turn around and walk in the other direction,
it's gonna flag the alarm.
- [Jason] The newest usage of artificial intelligence
is a camera that will train itself
to find unusual movements.
- The camera will learn the normal motion activities
in a scene over about a week or two weeks,
and after that learning period,
it will just automatically create an event and flag you
when something happens that's unusual for that scene.
- [Jason] The system has flagged people walking their dogs
as unusual motion detection at night.
- There is zero human input in this situation.
You basically turn it on, the camera learns
and does everything automatically.
- It sounds to me like this is in more places
than most people realize.
- Well, it is an active and live product today.
It is installed in many places across the U.S.,
most definitely.
- So there are real-time crime centers--
- We asked Andrew Sinclair,
general manager of Motorola Solutions,
about the privacy implications of the technology.
If something doesn't sem right, say something to the police.
Are cameras able to do the same kind of thing?
- There's things like anomaly detections.
- But what about someone acting suspiciously?
- Well, people's behavior is difficult,
'cause machine learning is not on that level yet.
But if a crowd suddenly appears at an intersection
where we wouldn't normally see a crowd
appearing at an intersection.
- I mean, the critics out there are saying
this is becoming surveillance on steroids,
that we're creating a surveillance state.
What do you say to that?
- I think it's important that we have the critics
out there saying that, because we have to make sure
that the right checks and balances are put in the space.
But you have to make sure that that same technology
can't be used to track your ex-girlfriend.
- One of these critics is Dave Maass.
He's an investigative researcher
at the Electronic Frontier Foundation,
a non-for-profit that defends civil liberties
in the digital age.
Maass is skeptical that the police can police itself.
- One of the things that worries me
about the future of policing is how automated it's becoming.
- [Jason] Maass and other experts are concerned
that these unsupervised systems make calls
on who belongs and who does not,
potentially amplifying racial bias.
For example, a camera might alert police officers
to a perceived motion irregularity it detected,
like an African American man walking through
a predominantly white neighborhood
with very little foot traffic.
They also point to the fact that every police encounter
is potentially dangerous, and these kinds of alerts
could lead to unreasonable stops and searches by the police.
- Artificial intelligence is becoming
a sort of black box with law enforcement,
where we don't necessarily know
where they're using it and how they're using it.
- [Jason] The next generation of body-worn cameras
will have AI-based search capabilities already built in.
- I'm looking for a person in a red shirt.
- [Virtual Assistant] Are you looking for person
in the red shirt?
- [Designer] Yes.
- [Jason] This way, all body-worn cameras in one city
could be searching for a man in a red shirt
or a missing child in real time.
- And that's what's concerning,
is this mass surveillance on society
and what kind of impact that will have on society
in terms of self-censorship, the ability to organize.
And I don't want to end up in a version of America
where people are afraid to do things in life
because of being watched all the time.

Share Video:

Embed Video: