Navigating the AI Frontier 2024

Evolution of AI agents’ capabilities FIGURE 2: TABLE 1:Technology trendDeterministic Non-deterministic Current state Agent type Simple reflex Condition– action rulesInternal model of the environmentTransfer learning and reinforcement learningEvaluating scenarios to choose the best outcomeCollaborative methodologies that represent the current state of the artBasic anti-virus softwareSmart thermostatAdvanced chess AIAutonomous drivingSmart city traffic plannerModel-based Goal-based Utility-basedMachine learning and deep learning techniquesLarge modelsMulti-agent systems Future type Agent examples Key characteristics Source: World Economic Forum Types of AI agents 2.2 This section outlines different types of AI agent and traces their evolution, highlighting the key technological advances that have supported their development. AI agents can be considered as either deterministic or non-deterministic, based on their defining characteristics, which are outlined below. Defining characteristics of deterministic and non-deterministic AI agents Deterministic AI agents Non-deterministic AI agents Rule-based: operate with fixed rules and logic, meaning the same input will always produce the same output. Data-driven and probabilistic: make decisions based on statistical patterns in data, with outcomes that are not fixed but instead are probabilistic. Predictable behaviour: the decision-making process is transparent and consistent, which makes the outcomes predictable. Flexible and adaptive: able to learn from data, adapt to new situations and handle uncertainty, often resulting in varied outcomes for similar inputs. Limited adaptability: these systems cannot learn from new data or adjust to changes; they follow only predefined paths.Complex decision-making: use algorithms that factor in probabilities, randomness or other non- deterministic elements, allowing for more nuanced and complex behaviours. Navigating the AI Frontier: A Primer on the Evolution and Impact of AI Agents 10