For decades, we’ve debated how emergency response units should be identified in a specific format when communicating on land mobile radio (LMR). Depending on a jurisdiction’s colloquialisms, it can be listed in many possible ways:
- Apparatus type (engine, truck/ladder, rescue, squad).
- Station number.
- Multidigit system guided by a state, federal, or international body with jurisdictional authority or “say-so.”
A simple Internet search of “how fire apparatus are numbered” produced about 72,100,000 results on the subject in less than a half-second. At the same time, we continue using the radio call signs as they are configured in the United States and worldwide.
- Communications Broke Down: An Excuse for a Serious Problem
- Understanding Emergency Communications
- Firefighter Basics: Fireground Communications
We have new language models to consider, as well as machine-made systems that adjust how business is done. Machine learning; artificial intelligence; and Natural Language Processing (NLP), the interdisciplinary subfield of computer science and linguistics, will have a lasting effect on public safety communication at a scale far beyond unit numbering.
Technology is advancing faster than anyone thought possible, yet digital tools have become agnostic and ubiquitous across the public safety domains. From “spell-checker” on text messaging to advanced NLP tools, we have much to consider in the implementation, utilization, and governance of these tools.
This includes writing materials such as the following:
- Decision briefs.
- Board presentations.
- Safety bulletins.
- Patient care summaries.
- Business inspection reports.
Chat-GPT and other NLP tools are primarily concerned with allowing computers to support and manipulate speech. Using them involves processing natural language datasets, such as text or speech corpora, and using rule-based or probabilistic (i.e., statistical and, most recently, neural network-based) machine learning approaches. The goal is to have a computer capable of “understanding” the contents of documents, including the contextual nuances of the language within them. The technology can then accurately extract information and insights from the documents and categorize and organize them. These “tools” (with no small measure of ethical origin consideration) impact the industry at an ever-increasing rate and are driven by pervasive tech industry investments.
With the continued incorporation of AI within daily life for public use, companies like Amazon have focused on creating the Amazon Web Services (AWS) GovTech Accelerator for public safety. Generative AI can be incorporated into multiple governmental areas, such as judicial and public safety, yet it is still susceptible to biased considerations and cybersecurity threats. While cloud-based AWS already has a presence in gov tech (a Smart City program), it is now focused on aiding cybersecurity cloud-based startups with guidance, funding, and mentoring. It emphasizes the fact that the incorporation of tech into governmental spaces is seeing a rapid increase.
Around the world, businesses are preparing to incorporate AI into their companies to increase productivity. Now more than ever, AI is leaving its past role as autopilot and entering its role as copilot within public safety strategies. There is a common consensus that the digital debt of the enduring legacy of communication systems like LMRs, static and siloed data sets, and manual notetaking is costing employers and first-response organizations in moving into modern innovation solution sets.
While employees continue to address their at-home and personal digital debt, they are asking for similar improvements on the job. As a result, an alliance between the usage of AI and current employees is becoming much more apparent. Employees are interested in utilizing AI to lighten their workloads and increase productivity while reducing the potential risk of burnout.
As AI becomes more commonly used, private and public companies must begin training their employees to be skilled in working and communicating with this technology. By learning how to write quality prompts and check for bias, employees can effectively use AI to their advantage.
Generative AI
ChatGPT-4 received the following question: “What is generative AI, and how might it be used by firefighters, rescue personnel, and emergency medical services?”
And this was the reply: “Generative AI refers to a category of artificial intelligence that can generate new data that resembles some form of input data. Unlike discriminative models, which classify or identify input data, generative models can create data that is similar but not identical to what they have been trained on. For example, generative models can produce realistic images, write text, compose music, and perform a variety of other creative or replicative tasks.”
Generative models include the following (and more):
- Generative adversarial networks (GANs): Primarily used for generating realistic images, GANs consist of a generator and a discriminator that work against each other.
- Variational autoencoders (VAEs): Used for developing new data. (Training data is a good example of this.) You might also use VAEs for natural language tasks such as text generation, translation, and summarization.
- Recurrent neural networks (RNNs): Previously famous for sequence generation tasks like text and music, though somewhat superseded by transformer-based models. Generation Z first responders may be more familiar with this one.
We as an industry—individually and collectively—continue to solve age-old challenges in two-way voice communications. While new words and phrases abound, some will have a positive economic impact on jurisdictions. These include the following:
- Unit number and the effects of digital debt on our business model.
- Internal customers.
- External customers.
- Credibility with enterprise.
- Efforts on the Hill.
Even so, one thing is sure: Our “language” is becoming digitally modernized.
Considerations as We Advance
In the epoch of digital transformation, the public safety domain is not exempt from the sweeping winds of change. Within the first half of 2024, 30,000 attendees gathered for the Microsoft GenAI Summit and 36,000 gathered for FDIC International. Enterprise investments are allowing for the integration of artificial intelligence, especially NLP and generative AI, to present a revolutionary approach to emergency response’s traditional methods and linguistics. While the debate on how to best identify responding units persists, the undeniably transformative potential of AI opens myriad possibilities, from advanced training simulations to predictive maintenance. With entities including Microsoft, Google, and AWS stepping into the fray to bolster public safety’s transition into the digital age, it is undeniable that the emergency response lexicon is on the cusp of a significant paradigm shift.
How Generative AI Can Be Useful to Firefighters, Rescue Personnel, and Emergency Medical Services
- Resource allocation: Generative models can predict the spread of a fire or other disaster scenarios based on existing conditions, helping commanders decide where to allocate resources more effectively.
- Simulation and training: Generative AI can produce realistic emergency scenarios for training purposes. For instance, it could generate a variety of simulated situations that firefighters and rescue personnel could use to improve their skills.
- Automated reporting and documentation: Text-generating AI can assist emergency services in quickly generating reports or documentation, allowing personnel to focus on critical tasks.
- Decision-making support: Generative AI can help generate multiple possible action plans in emergency scenarios, which decision makers can then evaluate for effectiveness. It can also suggest potential outcomes based on historical data.
- Communication: In situations where communication is difficult, generative AI can help create clear, concise, and practical messages or instructions for teams in the field.
- Patient care: For EMS, generative models can suggest likely diagnoses and treatment plans based on symptoms and other patient data, subject to confirmation by qualified medical personnel.
- Mental health support: Generative AI can create realistic simulations that help train emergency responders in high-stress or emotionally challenging situations.
- Predictive maintenance: Generative models can predict when equipment will likely fail, ensuring that vital tools and vehicles are in working condition.
- Data augmentation: Generative AI can produce synthetic data to augment real-world data. This helps train other machine-learning models that can assist in various tasks like object recognition, voice recognition, and more.
- Disaster response: Generative models can simulate the outcomes of various strategies (like controlled burns or building levees) under different conditions, helping emergency services prepare for natural disasters.
While generative AI has significant potential, it must be used responsibly and ethically, especially in high-stakes environments like emergency response. Proper validation, human oversight, and ethical considerations are paramount.
Using prompts in NLP and other generative AI systems is an evolving skill. Voice-to-text and text-to-speech tools, as well as image generation, help. And one day soon, brain-computer interface (BCI) will be a next step, at least for the brave and curious.
As with all advancements, we have inherent challenges to surmount, including cybersecurity threats, potential biases, and the steep learning curve associated with the integration of AI. Furthermore, the human element remains irreplaceable. Machines have yet to replicate the human mind’s empathy, intuition, and adaptability. Thus, the fusion of AI and human expertise provides a potent combination for the future. As AI assumes the role of the copilot, aiding and enhancing human decision making, it becomes crucial for public and private organizations to invest in training and awareness, ensuring a symbiotic relationship between humans and technology.
What Is AUDREY?
One of the most inspiring tools in the digital-decision domain was built at NASA’s Jet Propulsion Lab. At its core, the Assistant for Understanding Data through Reasoning, Extraction, and Synthesis (AUDREY) contains a revolutionary, domain-independent reasoning system capable of simultaneously performing inference and learning in real time.
With funding from the Department of Homeland Security Science and Technology Directorate, AUDREY can retrieve crucial data from disparate sources, dispatch analytics to extract information, and apply “humanlike reasoning” to synthesize and deliver essential insight to the first responders. AUDREY can act as an autonomous assistant, serving individuals and groups tasked with incident response. When time is fleeting and lives are at stake, the future of machine-assisted decision-making support will be a subject for the current and next millennia.
The unit’s call sign is a minimally complicated point in the context of emergency response linguistics. The advent of technologies like Chat-GPT, Gemini (formerly Bard), and BERT (short for Bidirectional Encoder Representations from Transformers) is a machine-learning model for NLP that’s impacting the language of the industry. It is as exciting as it is challenging.
Advanced AI tools signify the ushering in of a new era, where emergency response standards evolve from age-old analog methods to digitally enhanced systems, optimizing the efficiency and efficacy of public safety operations. However, as the industry embarks on this transformation journey, a balanced, ethical, and cautious approach will be pivotal in leveraging the full potential of AI while preserving the integrity and core values of emergency response.
On the other hand, multimodal large language models (LLMs) work with text and other forms of data, such as images, audio, and video (such as PaLM-E). These models convert text and other data types into a common encoding space, which means they can process all types of data using the same mechanism.
Lastly, members of the industry and those they serve will want to be looking for what comes next in moving voice from LMR to mission-critical push to talk (often seen as MCPTT) and moving from digital verbal debt communication to visual communications, like the connected fire apparatus image in this article, the depiction of which was generated by the OpenAI platform DALL-E (photo 1).
1. Photo by author, with AI assistance from DALL-E, using the prompt “Create an image of a smart fire engine, using the following: NextGen first response systems, remote sensing, biometrics, artificial intelligence, IoT, and related lifesaving technological solutions.”
REFERENCES
“Assistant for Understanding Data Through Reasoning, Extraction and Synthesis,” Department of Homeland Security. bit.ly/4bNU05R.
“AUDREY Watches Out for First Responders: An AI Case Study,” GovExec.com. bit.ly/3UWbypl.
FutureAnalytica, “Ethics in Natural Language Processing,” Medium, 11 February 2022. bit.ly/4brD70U.
Islam, A., “Multimodal Language Models: The Future of Artificial Intelligence (AI),” MarkTechPost, 19 July 2023. bit.ly/4dPi4XN.
McKinzie, K. “The Future of Artificial Intelligence in Firefighting,” Fire Engineering: Firefighter Training and Fire Service News, Rescue, 25 October 2018. bit.ly/3yxqW47.
Miller, B., “Connecting Minds to Keep the Peace: The Potential of Brain-Computer Interfaces in Law Enforcement.” Police1, 2 September 2023. bit.ly/4dW3rBV.
Rueter, T., “Amazon Launches Accelerator for Public Safety Tech.” GovTech. 7 June 2023. bit.ly/3KbuxaC.
Wegner, P., “What CEOS talked about in Q2/2023: Generative AI applications, bank troubles, uncertain economy.” IoT Analytics, 1 August 2023. bit.ly/3yxoDht.
“Will AI Fix Work?” Work Trend Index Annual Report, 9 May 2023. bit.ly/3WT57pW.
KIRK McKINZIE is a 35-year fire service veteran, a retired Consumnes (CA) Fire Department captain, and a first responder technologist. He is also an advisor, speaker, and author to government, enterprises, academia, startups, and operators focused on emergency response digitization and innovation for safer and more secure communities. He has taught at FDIC and his work has appeared in Fire Engineering and Fire Apparatus & Emergency Equipment.
