Check out the episode:

You can find the shownotes through this link.

---

Are you interested in urban progress with artificial intelligence?

---

Our summary today works with the white paper titled Agentic AI: The future is now from 2025, by VANTIQ.

This is a great preparation to our next interview with Nick Bray in episode 304 talking about how open architecture systems and artificial intelligence are changing our lives.

Since we are investigating the future of cities, I thought it would be interesting to see the rise of agentic AIs and their potential to revolutionise industries. This white paper introduces AI applications for healthcare, smart cities, public safety and other sectors to create better future for cities.

[intro music]

---

Welcome to today’s What is The Future For Cities podcast and its Research episode; my name is Fanni, and today we will introduce a research by summarising it. The episode really is just a short summary of the original investigation, and, in case it is interesting enough, I would encourage everyone to check out the whole documentation. This conversation was produced and generated with Notebook LM as two hosts dissecting the whole research.

---

[music]

Speaker 1: Today we’re going deep on AI, but not just the AI you’re used to.

Speaker 2: No, not just chatbots and image generators this time.

Speaker 1: We’re talking about AI that can act independently, AI agents.

Speaker 2: Yeah, it can be a little mind bending to think about, but AI agents are already out there, changing how things work in a lot of industries.

Speaker 1: Maybe let’s start with the basics for our listeners. What exactly is an AI agent? What makes it different from, say, just a really complex algorithm?

Speaker 2: At its core, an AI agent has a level of autonomy that sets it apart. Autonomy, okay. It means they can perceive their environment. That could be through sensors, data feeds, anything, really. And then, here’s the key, they use that information to make decisions, achieve goals.

Speaker 1: So it’s not just about processing data, it’s about actually acting on it.

Speaker 2: Exactly. An AI agent can analyse a situation and then take steps to reach a specific outcome.

Speaker 1: That’s pretty wild when you think about it. Like, it has a purpose, a goal.

Speaker 2: And that goal driven behaviour is another big part of what defines an AI agent. They’re not just reacting randomly, they’re designed with specific objectives in mind.

Speaker 1: Okay, so I’m starting to get the picture, but I’m guessing there are different levels of AI agents out there, right? Not all agents are created equal?

Speaker 2: Absolutely. There are different types of AI agents, each with its own strengths and limitations.

Speaker 1: Okay, I’m all ears. Hit me with the AI agent taxonomy.

Speaker 2: One way to categorize them is to think about how they learn and adapt. The simplest type is what we call a reactive agent.

Speaker 1: Reactive, like a knee jerk reaction?

Speaker 2: Kind of, yeah. They operate purely based on the current situation. Responding to immediate inputs without considering past experiences or potential consequences.

Speaker 1: So a bit like a thermostat that just adjusts the temperature based on the current reading. No complex learning involved.

Speaker 2: Exactly. They’re great for simple tasks, but their lack of memory or predictive ability limits what they can do.

Speaker 1: So what’s the next step up on the AI agent ladder?

Speaker 2: We call those model based agents. These guys are a bit more sophisticated. They actually have an internal model of their environment.

Speaker 1: So they can actually think ahead.

Speaker 2: In a way, yeah. They can anticipate the consequences of their actions and plan accordingly.

Speaker 1: That’s a big jump in capability. So instead of just reacting to the white now, they’re thinking, if I do this, then what?

Speaker 2: Exactly. It allows them to be much more strategic, adapt to changing circumstances.

Speaker 1: Can you give me a real world example of that? How a model based agent might operate?

Speaker 2: Sure. Imagine a financial training system that uses a model of the stock market. It can predict future trends and make smarter investment decisions based on that model.

Speaker 1: Okay, so it’s not just about reacting to price changes, it’s actually trying to anticipate what’s going to happen next.

Speaker 2: Precisely. Now, the most advanced type of AI agent is the learning agent. These guys are constantly evolving.

Speaker 1: Oh, this is where it gets really interesting.

Speaker 2: Yeah, they can actually improve their performance over time through experience. They analyse feedback, data, even their own mistakes to refine their strategies, become more effective at hitting their goals.

Speaker 1: They’re like that friend who always seems to be learning from their mistakes and getting better at everything they do.

Speaker 2: That’s a good analogy, except without the human emotions and all that.

Speaker 1: So we’ve got these different types of A. I. agents out there. Why are they such a big deal? Why should people care about this whole evolution of A. I.?

Speaker 2: That’s a great question. A. I. agents have the potential to tackle some of the most complex challenges we face across pretty much every industry.

Speaker 1: Really? That big of a deal?

Speaker 2: Oh, yeah. They can analyse huge data sets, see patterns that humans would miss, make decisions in real time. Address problems as they pop up.

Speaker 1: Okay, so we’re talking about solving problems that are just too complex for humans to handle on their own.

Speaker 2: Think about something like traffic management in a busy city. An AI agent could analyse data from traffic cameras, GPS systems, even social media to get a handle on what’s happening.

Speaker 1: Wait, so social media too?

Speaker 2: Sure, it can pick up on accidents, road closures, anything that might be impacting traffic flow. And then the agent can use all of that information to optimize traffic light timing, suggest alternate routes, keep things moving smoothly.

Speaker 1: That’s incredible. So it’s not just about making things a little bit more efficient. It’s about solving problems that were just impossible

Speaker 2: AI agents being used in manufacturing. Healthcare, finance, all sorts of fields.

Speaker 1: It’s starting to click for me why everyone is so excited about this technology.

Speaker 2: Oh, it’s early days, but the possibilities are vast.

We’ve only scratched the surface here.

Speaker 1: I’m definitely intrigued, but how does all of this work in practice? What are some real world examples that go beyond just the theory?

Speaker 2: So one area where AI agents are already making a big difference is in disaster response. High stakes situations where every second counts.

Speaker 1: Okay, I’m picturing something intense.

Speaker 2: Imagine a wildfire raging through a forest, threatening homes and communities. An AI agent, hooked up to a network of sensors and data feeds, could analyse real time information. Wind speed, terrain, fire spread patterns.

Speaker 1: So it’s like having an AI firefighter with a bird’s eye view of the entire situation, constantly analysing and predicting what the fire will do next.

Speaker 2: And it goes beyond just

Speaker 1: prediction.

Speaker 2: This AI agent could also help coordinate the deployment of firefighting resources, making sure firefighters and equipment are in the right place at the right time.

Speaker 1: That’s amazing. It sounds like AI agents could significantly improve how we respond to disasters, making those efforts more efficient and effective.

Speaker 2: Exactly. And that’s just one example. AI agents are being applied in a wide range of fields, from healthcare to finance, transportation, and beyond.

Speaker 1: Speaking of healthcare, you mentioned AI agents being used to personalize treatments earlier. It sounds futuristic, to be honest.

Speaker 2: It might sound like sci fi, but it’s closer than you think. Picture an AI agent that continuously analyses a patient’s medical history, genetic data, even real time vital signs.

Speaker 1: So it’s like having a personalized medical expert, always on call, making sure you get the best possible care.

Speaker 2: That’s the idea. It has the potential to revolutionize healthcare as we know it.

Speaker 1: Okay, colour me impressed. She also mentioned finance. How are A. I. agents changing things up in that world?

Speaker 2: Fraud detection is a big one. These A. I. agents can sift through mountains of financial data, transactions, looking for patterns that might indicate something fishy is going on.

Speaker 1: So it’s like having an army of AI detectives constantly patrolling the financial system, making sure everything’s above board.

Speaker 2: They can spot those subtle red flags that humans might miss. And they can react in real time, stopping fraud before it even happens.

Speaker 1: That’s impressive. But all this talk about super smart A. I. agents making decisions, It makes you wonder, are there any downsides? Any risks we need to consider?

Speaker 2: It’s a valid question. AI agents are incredibly powerful tools. But we need to make sure we’re using them responsibly. That brings up a whole bunch of ethical considerations.

Speaker 1: Are we building these things with the right values in mind? Making sure they’re used for good, not for harm?

Speaker 2: Exactly. We have to make sure that AI agents are aligned with human values and goals. Their actions need to be transparent and accountable. We can’t just let them run wild. We need to think about bias, privacy, security, the whole nine yards.

Speaker 1: So it’s not just about making AI agents as smart as possible. It’s about building them in a way that benefits humanity.

Speaker 2: 100%. The goal isn’t to replace humans. It’s to empower them, to help us solve problems and create a better future.

Speaker 1: It’s a lot to think about.

Speaker 2: Yeah, AI agents are powerful tools, but they come with big responsibilities.

Speaker 1: I’m ready to go even deeper. Let’s talk about the technical side. How are these complex AI agent systems actually built? What does it take to manage them effectively?

Speaker 2: It’s true that building AI agent systems can be a pretty daunting task, especially as they get more complex and you start dealing with lots of agents working together. That’s where the idea of an AI agent platform comes in.

Speaker 1: You mentioned AI agent platforms earlier, but I’m still a bit fuzzy on what exactly they are. Are they like the operating system for AI agents, the foundation on which everything else is built?

Speaker 2: Yeah, that’s a good way to think about it. An AI agent platform provides the tools and infrastructure you need to create, deploy, and manage all those intelligent agents. It’s what makes the whole thing work.

Speaker 1: Okay, so if a company wants to start building and using AI agents, they’re probably going to need a good platform. But there must be a ton of platforms out there. What are some of the key features that make a platform stand out? What should companies be looking for?

Speaker 2: One of the most important things is real time data processing. AI agents need to be able to make decisions based on the most up to date information. So the platform needs to be able to handle a constant stream of data and process it quickly.

Speaker 1: So it’s not just about storing data, it’s about Being able to analyse it and act on it in real time.

Speaker 2: Exactly. Think about an AI agent managing a fleet of self driving delivery vehicles. It needs to know about traffic conditions, weather changes, new orders coming in. All of that has to be processed and factored into the decision making process in real time.

Speaker 1: That’s a perfect example. It makes sense that speed and adaptability are key. What other features are essential for an AI agent platform?

Speaker 2: Another crucial element is asynchronous execution. This basically means that the platform should allow AI agents to handle multiple tasks at the same time.

Speaker 1: Okay, so it’s like multitasking for AI agents.

Speaker 2: Pretty much. If an agent is managing security for a building, it needs to be able to monitor security cameras, analyse access logs, and respond to potential threats all at the same time. Asynchronous execution makes that possible.

Speaker 1: So we got real time data processing and multitasking. What else is on the must have list for a solid AI agent platform?

Speaker 2: We’ve talked quite a bit about large language models, LLMs. Integrating those into the platform is becoming increasingly important.

Speaker 1: Because LLMs can give AI agents access to so much information and make them better at understanding and generating human language.

Speaker 2: Exactly. LLMs can help agents with tasks like customer service, content creation, even complex decision making. They’re incredibly powerful tools.

Speaker 1: I can see why integrating LLMs into the platform would be a game changer. Anything else?

Speaker 2: Another really important factor is collaboration between AI agents and humans. The platform should provide tools that allow humans to set goals, define parameters, and monitor what the agents are doing. It’s about having a system where humans and AI can work together effectively.

Speaker 1: That makes a lot of sense. You don’t want to just set these AI agents loose and hope for the best. You want to make sure they’re working towards the goals that humans have set.

Speaker 2: Absolutely. And as you start to have more and more AI agents working together in a system, you need a way to coordinate all of them. That’s where orchestration

Speaker 1: comes in. You mentioned orchestration earlier. It’s like conducting an orchestra, but with AI agents instead of musicians.

Speaker 2: A good AI agent platform needs to have strong orchestration capabilities. It needs to be able to manage the communication between agents, allocate resources, and make. sure they’re all working together smoothly towards the same overall goals.

---

[music]

What is the future for cities podcast?

---

Episode and transcript generated with ⁠⁠Descript⁠⁠ assistance (⁠⁠affiliate link⁠⁠).