In 2004, Bill Gates famously predicted that “Two years from now, spam will be solved”. Clearly, he missed the mark. Predicting the future is not just challenging, it’s virtually impossible especially in the field of technology. Artificial intelligence (AI) has been called our last invention. Undoubtedly, it will transform our society. What that transformation looks like exactly is yet to be determined. Regardless, we should be incredibly skeptical of absolute statements regarding the capabilities and future of artificial intelligence. How prepared is your department and personnel for this rapidly evolving technology? We simply cannot afford to bury our heads in the sand and ignore the impact this will have on our communities. Leaders must fundamentally understand what AI is, how it works, and where we can leverage its power for the betterment of our departments and communities.

The latest studies show that the average American uses AI 25-30 times a day. Most of us don’t realize it. Do you use facial recognition to unlock your phone- that’s AI. Your text messages, emails, calendar appointments, and Google searches all have artificial intelligence embedded within them. AI has slowly immersed itself in our daily lives.

The term “Artificial Intelligence” was first coined by John McCarthy in 1956 at Dartmouth University. That was 70 years ago, a timeline that underscores the difficulties scientists have faced in replicating the human brain. However, recent advancements in computer science including algorithmic innovations, have moved us beyond simple programming. We have now entered what is known as Generative Artificial Intelligence (GenAI). In short, machines are now able to generate new content. We have achieved AI through what is called machine learning.

But what exactly is artificial intelligence? I define it as:

A machine’s ability to perform the cognitive functions we typically associate with human intelligence and problem-solving capabilities, without explicitly being programmed to do so.

“Without explicitly being programmed to do so” is the key. Never has a technology been able to create new content and make such accurate predictions. GenAI is achieved through Large Language Models (LLM’s). Without becoming too technical, scientists have created an artificial neural network (ANN) to replicate how we believe the human brain works. Within ANN, there are three layers- the input layer (your data), the hidden layer, and the output- the results. Each of these layers contain nodes which are interconnected with parameters. Parameters can be adjusted by their weights or biases, which controls the “temperature” or creativity of the model.

I’ve been fortunate enough to collaborate with software engineers who all independently say “We don’t truly understand why this works”. Collectively they agree the latest AI models are only generating the next word that it (the machine) believes makes the most sense based on its training. As I’m sure you’ve experienced, it’s not always right. The technical term for this is hallucination. The model has made an inaccurate prediction of what the next word should be.

AI is an incredibly broad term. It can be challenging for some to wrap their heads around this idea. AI was first introduced to the masses in 2011 with Siri embedded on the iPhone. This is referred to as Traditional AI. The next evolution of AI is called Generative AI. Generative AI is where we are today. Next is what is called AGI- Artificial General Intelligence. This is defined as replicating human cognitive ability across a wide range of tasks at a Nobel Laureate level. Some of the characteristics of AGI are versatility, adaptability, self-improvement and general understanding. AGI what every tech company is aggressively pursuing. Hundreds of billions of dollars are being invested into achieving AGI. 

We are seeing exponential growth with AI. For comparison, if a human did nothing but read 24/7 for their entire life they would consume 8 billion words. The latest AI models use 8 trillion words in a single month of training. That is 1,000 times more in one month than what a human could do over the course of their lifetime. An example of the exponential growth of this technology is seen when we examine the scale in which these models are being trained. For example, ChatGPT’s first model was trained with 117 million parameters. ChatGPT’s 4^th model was trained with 1.7 trillion parameters, which only took three years. ChatGPT 4 has an IQ of 155. For reference, Einstein’s IQ was 160.

Now that you understand the basic framework of how AI works, we must examine how it will impact our agencies. We utilize a variety of tools to achieve our mission of public safety. Artificial intelligence is just that, a tool. Do you require your officers/staff to receive training before permitting the use of a certain tool? Obviously, the answer is yes. We would never deploy a piece of equipment and hope for the best. We spend an incredible amount of time and money ensuring our officers can effectively use these tools to accomplish the mission of public safety. Artificial intelligence is no different. It is a tool that demands a fundamental understanding so its power can be leveraged to your advantage while mitigating the risks associated with it. What training has your organization received on AI?

When I first began speaking on artificial intelligence, and understood its power, I quickly realized our department didn’t have a policy. I then did as any Chief would and called surrounding jurisdictions asking for a copy of their policy, as there is no reason to reinvent the wheel. I called no less than 20 agencies asking for a copy of their policy. After gathering all their policies and examining each closely, I realized they all had one thing in common- not a single agency had a policy on artificial intelligence. We then drafted a policy which governs the use of AI within our agency. It limits the use of what it can be used for and among other things, requires a human to be involved before any adverse action can be taken against someone.

Progressive police departments live and die by policy. Among other things, policies provide agencies with a roadmap of how to handle various scenarios to protect the officers, the citizens and the department. Does your agency have a policy on grooming standards? If so, why? We want to project a professional image to our citizens because this is important to us. Do you have a policy governing the use of force? Of course, we want to ensure compliance with the least amount of force necessary. With that in mind, do you have a policy to regulate the most powerful tool within your agency?

Are your officers using AI to write their reports, or is it being used to make their reports better, more articulate? Is it being used as a tool for predictive policing, or facial recognition? Is it CJIS compliant? Policies will help mitigate your liability by giving your officers and staff guidance on acceptable practices.  As AI becomes increasingly prevalent, we should expect there to be more questions raised about the use of this technology.

Currently there are several lawsuits filed against police departments across the country because of the actions they have taken based upon the results of AI. Whether it’s facial recognition, predictive policing, or generative report writing, agencies are facing significant scrutiny. Blindly relying on AI to make decisions which can adversely impact citizens is a serious issue. While significant advancements have been made, AI is not a silver bullet. When these systems hallucinate, we must have policies in place to act as guardrails ensuring equitable treatment of all our citizens. Innocent people have been arrested based solely on the results of AI. As leaders in this profession, I trust you are as outraged as I am. I don’t fully blame AI, but rather the lack of policies in conjunction with poor police work.

Artificial intelligence is undoubtedly the “smartest” member of your team. Its capabilities are truly amazing. Companies are aggressively pushing their “latest and greatest” product, which touts the power of AI and promises to solve all your department’s problems. Companies are rushing to market with products that have not been properly tested. The District Attorney in Seattle Washington refuses to prosecute cases where AI was used to generate a report. They recognized the dangers of hallucinations and in the judicial system we simply cannot afford to make mistakes. The Anchorage Police Department conducted a study and found that using AI to generate reports did not save the officers any time, due to the number of errors or hallucinations it produced.

Before engaging with companies who utilize the power of AI within their products, you first must understand how their AI works. Someone within your organization needs to be technically savvy enough to ask the right questions. Was their AI system independently tested and if so, what were the results. Was the training supervised, unsupervised, or reenforced. What were the biases used? We must be incredibly meticulous in reviewing a company’s artificial intelligence program before signing a contract.   

I strongly encourage you to audit your department and determine which systems are using AI. I believe you will be shocked at the number of programs which have AI imbedded within them. Once those programs are identified, examine exactly how AI is being used within that program. Is the system making predictions, producing information which can lead to an arrest, or a search warrant? If so, are there guardrails in place to mitigate your liability?

As leaders we have a choice, to either evolve or be left behind. This profession has and will continue to evolve. Artificial intelligence is here to stay. We simply can’t afford to ignore this technology given all the good that it is capable of as well remaining cognizant of the inherent dangers that will accompany it. Artificial intelligence is the most powerful piece of technology we’ve ever invented, and we must establish a robust framework to ensure it is used fairly and judiciously.