PCG logo
Article

2024: The Year of AI Agents

image-fc21aa016ff3

The year 2024 is all about AI agents. But what exactly are AI agents? To understand this, we first need to take a look at developments in the field of generative AI. One of the most significant trends is the shift from monolithic models to complex, composite AI systems.

image-4b1bb4cbf0d5

From Monolithic to Modular Systems

image-351f67d531c7

Traditional AI models are based on the data they were trained on. This means they only have knowledge of the world that is available to them and are often used for specific tasks. Additionally, these models are difficult to adapt. Even if it is possible to improve or fine-tune a model, it requires a significant amount of data and resources.

For example: Let’s say I want to know how many vacation days I still have left. If I ask this question to a generative model, the answer will likely be incorrect, as the model doesn’t have personal information about me. This limitation can only be overcome by embedding the model into a larger system that can access the necessary databases.

Composite AI Systems: The Future

This is where composite AI systems come into play. These systems consist of multiple components that work together to solve complex tasks. In our vacation example, this means that the AI model wouldn't just give a generic answer, but also access a database where my personal vacation data is stored. The model would then create a query, retrieve the data from the database, and generate the correct response: "You have 10 vacation days left."

image-3b5e47b75c45

This principle makes it much easier to adapt such systems compared to adapting individual models because it is faster and more flexible. A well-known example of such a system is "Retrieval-Augmented Generation" (RAG), where models are combined with a search function to retrieve specific information from external sources.

The Evolution: AI Agents

This is where AI agents come into play. A significant breakthrough in AI is the ability of large language models (LLMs) to analyze and plan for complex problems. Essentially, an AI agent is such a model that is not only capable of processing data but also of acting strategically, using external tools, and adapting to various challenges.

An agent can not only access databases but also utilize external tools to perform calculations, search the web, or run programs. So, in our vacation scenario, an AI agent could predict the weather, calculate the amount of sunscreen needed based on sun exposure, and even recommend the number of sunscreen bottles to bring.

What Makes AI Agents Special?

Three key abilities differentiate AI agents from traditional AI systems:

  1. Reasoning and Planning: Agents can think through complex problems, break them down into smaller steps, and solve them systematically. Instead of providing a simple, possibly incorrect answer, they follow a path of careful planning and testing.
  2. Action Capability: AI agents can use external tools to complete tasks, whether it’s querying a database, performing a web search, or using a calculator for mathematical calculations.
  3. Memory: Agents can remember past interactions and conversations, allowing them to work contextually and give personalized responses.

One of the best-known configurations for AI agents is the so-called ReACT model, which combines reasoning and action capabilities. Agents working under this principle don’t just think; they also actively act, call external tools, and correct themselves if mistakes occur.

An Example: My Vacation in Thailand

image-7fef7eac0edd

Let’s imagine a concrete scenario: I’m planning a vacation in Thailand and want to know how many bottles of sunscreen I should bring. An AI agent could break down this complex task into several steps: First, it accesses my vacation data to determine the number of remaining vacation days. Then it checks the weather forecast for Thailand to calculate the hours of sunlight. Next, it researches the recommended amount of sunscreen per hour and finally performs a calculation to determine how many bottles I need.

Conclusion: The Future of AI

2024 will be the year of AI agents, as they represent a new level of autonomy and problem-solving. While simple, programmatic systems will still make sense for narrowly defined problems, AI agents offer a solution for more complex tasks through their flexibility and adaptability.

The Public Cloud Group is here to accompany you on this exciting journey. We help you find the right solution for your specific challenges, whether through the use of composite AI systems or the integration of agent-based intelligence. Our experts support you in developing and implementing these innovative technologies to make your business processes more efficient, flexible, and future-proof.

Don’t hesitate to contact us – we are your partners on the path to the world of AI agents.


Services Used

Continue Reading

Article
AI
Unlocking Business Growth with Prompt Engineering Techniques

Discover how Prompt Engineering drives business growth and innovation.

Learn more
Article
Leveraging Public Cloud Platforms for GenAI

Discover how AWS, Azure, and Google Cloud help choose and deploy AI models, ensuring flexibility and scalability for AI success.

Learn more
Article
A Framework for Selecting the Right AI Foundation Model

Learn how to select the best AI foundation model for your specific needs with a step-by-step framework, evaluation factors, and a multi-model approach.

Learn more
Article
Foundation models in AI: mastering the challenges of selection

Explore foundation models in AI, their challenges, and examples. Learn how to select the right model for your needs.

Learn more
See all

Let's work together

United Kingdom
Arrow Down