In Part 1 of this interview, Veritone’s Rob Gerber, a former detective at Antioch Police Department in Northern California, discussed how he got into the world of investigations, the largest challenges he faced and how they evolved with the growth of digital evidence sources, and how low staffing and retention rates impacted investigations across the department.
If you missed Veritone’s March 20 webinar discussing how AI can accelerate investigations, increase case clearance rates and help bring justice to those affected by crime, find it on-demand here.
Police1: What specific AI-driven technologies are now available to investigators, and how has that changed the workflow of traditional investigations?
Rob Gerber: The digital tools that are available have definitely improved as of late. There’s traditional things like cell phone extraction tools and cell phone record analysis tools and all those types of things that are really, really crucial, but [now we can allow] AI to come in and sift through data either from an analysis perspective, [or] maybe it’s combing through large data sets and making connections.
In my experience, the most important thing has been where it can review video or audio for you. The biggest time sink for most investigators is the large amounts of maybe 100, 200, 300 hours of video, after a homicide, you’ve got to review to find the one piece that matches another piece of the video or find the one car that makes it through all of the video.
That manual labor is really where AI can come in and speed things up. It can do that review process for you and bring to the top the actual piece of evidence that the investigator then can review.
I think you already answered my next question, but I’m going to ask it again, just in case you want to add anything to it about how AI has improved the efficiency of digital evidence review and analysis and other laborious tasks like audio/video redaction.
Gerber: The human still has to stay in the loop. The detective or the officer or the records personnel still has to do the manual work to make sure it’s correct and perfect for our citizens, but the AI can definitely drastically speed up that front end.
If the AI can find every face in a piece of video so that you can know which one to redact, and instead of drawing a box around those faces, you can just click those faces, you can speed up a redaction of a video by 80%. Same with reviewing … surveillance videos: If you’ve got 100 hours of video, and the AI can look for when the person goes from point A to point B and find you every moment in the videos that happens, now your detectives only have to review those 45 seconds instead of those 100 hours.
It really is speeding up those time-intensive or labor tasks – the manual labor tasks of digital evidence – allowing the detectives and analysts to really focus on what’s important, which is correlating that data and trying to solve a crime.
How can an AI-driven intelligent analytics solution like Veritone iDEMS ensure both security and scalability?
Gerber: Securitywise, you can make sure it’s compliant, and Veritone’s iDEMS [Intelligent Digital Evidence Management System] happens to be with CJIS – or, if it’s on the federal side, FedRAMP – in that it’s completely housed in a secure environment.
It’s not reaching out to open source, which is the real fear with AI these days. “Is it going to reach out to the internet, where it’s going to bounce your information out?” Something like Veritone and CJIS compliance allows it to all stay in a secure environment – a cloud environment that’s tried and true with multiple kinds of securities and fail-safes built in to make sure it’s secure and compliant with all the security needs of the public sector.
Then on the second portion of your question, making sure that it’s scalable. It’s really a matter of not having a hard limit on the amount of evidence or the speed of that evidence, and that’s allowed by using a large cloud.
So, in the case of iDEMS, it’s on Amazon AWS [Amazon Web Services]. It’s a huge cloud environment that’s built to scale with the largest companies in the world. So that allows you as a police department to have the resources you need when you have that high-profile homicide or have the resources you need when there’s a small residential burglary with no real victims, but you need to solve that crime for your citizens now. No matter the size of crime or the size of resource you need, it’s there for you.
Can you share an example of a case where limitations and the kinds of traditional investigative tools available to you at the time either significantly impacted the investigation or delayed justice? Or maybe, alternatively, any cases where, in retrospect, had you had AI tools available to you, it may have resulted in a faster or better outcome?
Gerber: Probably a year prior to my retirement, I worked a very high-profile homicide within the city of Antioch that turned into a large task force because the responsibles for that homicide ended up killing people throughout the Bay Area. There ended up being a wiretap. It was about a year [of] investigation.
And [with] all of that, we’ve now realized, if we inserted those cell phone downloads in the initial surveillance video that we had … into something like Veritone, where it can correlate the data and review the surveillance video, compare it to all of the images inside of a cell phone, we would have seen immediately that the same van was used on one of our initial crime scenes before he kept murdering people.
[It] was the same van that our suspect was trying to sell on social media that we had immediately. But we manually would have never made that connection. We didn’t make that connection until we were going to court a year later after the wiretap. That would have saved us 10-plus months of investigation, probably a couple hundred thousand dollars of the wiretap investigation, and all of those employees that were working on it.If we made that connection very early, we probably would have solved it within a month or two.
SUCCESSFUL OUTCOMES AND HOW AI CONTRIBUTES
That’s quite an impact that AI made in that case. So, tell me, how do you measure … success in your time as investigator?
Gerber: It’s really just [to be] able to solve crimes and bring justice to families. That’s the only level of success I ever had. It wasn’t necessarily case clearance rate or any specific number I could point to. It was being able to bring some resolution to those families that were victims. And being able to do that brings the reward, and it’s the reason most police officers and specific investigators do what they do. And if I can assist to make that easier or more efficient, we’ve got to utilize that to be able to bring more justice and more solves to those police departments.
And what does it mean to you personally to be able to bring that resolution to families?
Gerber: That’s why I was in law enforcement. I unfortunately had to retire out for other reasons. But during my time, that was the only reason I was doing what I was doing. The only reason I was pulling long hours and staying away from my own family was to bring that resolution for those families of those victims, specifically homicides. They – the victims – can’t speak for themselves, so you need to do that work for them and then … bring a solution.
It’s not bringing the loved one back but bringing an answer and a conclusion to that horrific piece of their life to them [that] made it all worth it. Now being able to at least bring some technology to the police departments to be able to help them do that better is great.
In your current role, do you ever encounter resistance from investigators or other law enforcement professionals who might be skeptical about AI-driven technologies? And, if so, what are those concerns that they have, and how do you address them?
Gerber: There’s definitely those that are resistant to change just in general. Also in law enforcement, specifically, people have been doing it a long time one way. Bringing in a new way of thinking or doing things is difficult. So, there’s always those that are just kind of resistant to change.
And then also some people are just worried about the security. They’re worried about data breaches and leaks and its ability to be a secure, reliable piece of information that they can then testify to. But in the end, this is the direction that technology is going, and there’s no way police departments can really keep up with the amount of crime that’s occurring and the staffing shortages and just the large amounts of digital evidence that every year we have more of.
They’re just going to have to understand the technology and try to apply it in a way that makes sense for them and their department.
MAKING THE MOST OF AI
What are some of the best practices that an agency should do if they want to implement AI technology into the workflows?
Gerber: Best practices really are making sure that the solution they choose is compatible with everything else that’s on board, everything else that they already utilize. Most departments are already utilizing certain technologies. They need to make sure that the tech that they choose next plays well with others. They need to make sure it’s in a secure environment, that it’s CJIS-compliant and it’s also compliant to their specific state laws.
Some technologies may be overreaching on something like facial recognition, and it’s not written or [correctly] created in a way that complies with their state laws. So they’re going to have to just make sure it complies with CJIS and their specific state laws.
Then just making sure there’s always that human in the loop. It’s not relying on AI to solve things. It’s not relying on AI to say, “That’s my suspect,” but instead produce leads that a detective can then utilize to solve the case.
What is your favorite feature with Veritone iDEMS?
Gerber: It’s probably two. It’s the two features I would have used the most as an investigator. And that’s one, reviewing surveillance video after every homicide. We would do a video canvass collect – a couple hundred hours of surveillance video from 20 to 30 cameras. Being able to have artificial intelligence review that [and] initially show me the actual pieces of evidence that are relevant and helpful and not have me or another analyst review that hundreds of hours of data … that would have been one of the biggest features.
The other is just the cell phones. I’ve spent months of my life staring at cell phones of criminals, and to be able to instead search through those, find just my suspect or just my victim or just the relevant information about the gun or about the drugs and specifically videos and pictures and texts – that would have saved me months of my life.
That’s the things I’m most excited to pass along to other investigators.
Veritone’s mission is “AI for good.” How do you feel Veritone is delivering on that mission?
Gerber: Really the goal of things is [to] provide AI in a way that’s trustworthy – that’s compliant with all regulations and truly secure. But really “AI for good” allows you to bring transparency to the public through things like redaction and those types of efforts, but also bring efficiency and empower those investigators and those officers to do more with less.
There’s just not enough cops. There’s not enough law enforcement. And there’s too many crimes and too much digital evidence to go through. If our AI technology can help us to solve more with less, to be more efficient, to give the tools to those investigators, to bring closure to families that are the victims or the victims of violent crimes … really the whole purpose of Veritone being in the public sector is providing those tools that allow the law enforcement to do more and provide more to the public that truly needs it.
So where do you see the future of AI-driven tools and in investigations or in law enforcement generally? And what do you think it’ll look like in five to 10 years?
Gerber: I don’t think it’s going to be “Minority Report.” I don’t think it’s going to be specifically finding criminals, but the more AI is developed in a great way… It’s really about providing efficiencies, being able to ask artificial intelligence to do the manual labor, the culling of data, the specific correlation of that data.
“This one tattoo: Where is this in all of my evidence? Who else has that tattoo?” Those types of things where manually it would take you hundreds of hours, but you can do quickly with artificial intelligence is really, from my perspective, the best way that we can utilize it while having a detective at the end of that process, verifying and using the data as he sees fit.
So, Rob, is there anything that I’ve forgotten to ask you that you want to add?
Gerber: I don’t think so. I’m just excited for more police departments to be able to utilize technology to speed up the workflows, make up for the lack of staffing and solve more crimes. AI’s, for my perspective, one of the best ways to do that. Veritone iDEMS happens to provide all of those tools that can really speed up the investigations. I’m just excited for other people to learn about that.
Note: The above conversation has been condensed and lightly edited for clarity.
