Having previously explored what an intelligent agent in AI is, we now turn our focus to how many types of agents are defined in artificial intelligence. Grouped by their range of capabilities and degree of perceived intelligence, there are five categories and each has the power to turn AI ideas into action.
Solely working on current perception, these agents do not consider the history of perception and that is why they are only successful when they have an environment that can be fully perceived. Befitting an agent with limited intelligence, they are limited in their capabilities. They are not adaptive to the environment and if something is not perceived in the current state, it will not be part of the action. Responses are essentially based on a user initiating an event and the agent referring to a list of pre-set rules and pre-programmed outcomes.
These agents have a key advantage over Simple Reflex Agents – they consider the history and thus can work in an environment that is not fully observed. Boasting a model and internal state allows the model to tell about the current state of the world and the internal state to tell about the current state based on the history of perception. While choosing an action in the same way as a reflex agent, they have a more comprehensive view of the environment.
As indicated by the name, these types of agents use goals to describe desirable capabilities and, in turn, can choose among various possibilities. An extension of Model-Based Agents, they choose the best action from the available options to reach the goal, with the decisions made by artificial intelligence. The fact these agents do make a choice means the process is referred is known as ‘searching and planning’ to make an action.
While similar to Goal-Based Agents, they boast the advantage of providing an extra utility measurement that rates potential scenarios on their desired results and then opts for an action that maximises the outcome. This also allows it the ability to trade-off different factors before making a decision. For example, a clothing store’s goal may be to make a profit on sales but the utility recognises customer satisfaction also needs to be considered when pursuing that profit. By setting the utility as a real number (eg: a scale of 1-10 of customer satisfaction), the Goal-Based Agent is capable of deciding real-world scenarios based on utility.
An additional learning element means these agents can gradually improve and become more knowledgeable about an environment over time. It does so by taking feedback from whatever actions it has performed and adapting accordingly. This process requires the Learning Agent to have four components – the learning element (which learns from experience); the critic (which is the feedback system); the performance element (which decides the external action that should be taken); and the problem generator (which is a feedback agent that keeps history and makes new suggestions).
The rise of artificial intelligence knows no bounds. From forecasts that the sector’s annual growth rate will be 33.2% between 2020 and 2027 to research showing 80% of retail executives expect their companies to adopt AI-powered intelligent automation by 2027, it is clear organisations yet to explore adopting AI strategies risk being left behind. Education is also essential and when it comes to AI, appreciating the important role played by intelligent agents is a crucial first step.
With successful automation taking time, extensive research and strategic assessment, consider these essential tips when preparing your own Robotic Process Automation journey.
Following the Contact Centre Week interview, Probe CX's CEO Andrew Hume discusses the work-from-home mobilisation of its global organisation and what it means for the future of work.
Do you know what PDDs and SDDs are in RPA? Here’s the difference between the two, as well as how they develop an automated process.
The finance and accounting sector is burdened by repetitive and time-consuming tasks, which is why robotic process automation is ideal...