The course "Laboratory of Artificial Intelligence Applied to Language" is part of the Cognitive Science of Communication and Action training activities.
The lab aims to provide students with a solid theoretical foundation of AI applied to language and a well-established ability to interact with Large Language Models (LLMs) through the most advanced prompt engineering techniques.
These objectives are achieved through two phases.
The first, more theoretical, involves teaching how neural networks work with a particular focus on Transformers, i.e., the architecture used in commonly used language models today such as ChatGPT.
The second, practical phase involves teaching prompt engineering techniques and the student's application, including experimental application, of these techniques in the various scenarios that will be presented.
Expected accomplishments
By the end of the course the student will have understood both theoretically and practically how these models work.
Theoretically he or she will have understood:
- how neural networks are similar to human functioning and how they differ;
- how neural networks and in particular language model neural networks (LLMs) work; and
- What prompt engineering techniques should be used depending on the context and in general how to relate to these models
- How they can be used in society, work, and their socio-economic implications.
Practically, the student will be able to:
- Write the most appropriate prompts for the required case and know how to get the most out of language models
- Understand the limitations of such models and how they can be used in a pipeline to create linguistic automations.
No prior knowledge of code is necessary
teacher profile teaching materials
These objectives will be pursued through two complementary phases.
The first phase, mainly theoretical, will focus on understanding how neural networks work and, in particular, on the Transformer architecture underlying modern Large Language Models. Students will examine the fundamental principles that enable these systems to process and generate natural language, as well as their main limitations and risks.
Topics covered will include, among others, bias, hallucinations, prompt injection, human-machine interaction, human-in-the-loop approaches, and the role of language in interaction with Artificial Intelligence systems. The social, economic, and ethical impacts arising from the widespread adoption of generative AI will also be discussed.
The second phase, mainly practical, will focus on prompt engineering techniques and their application across the different use cases presented during the course. Students will directly experiment with language models and use these tools to carry out exercises and create digital artifacts, customized assistants, and simple AI-based workflows.
# Learning Objectives
By the end of the laboratory, students will have acquired both a theoretical and practical understanding of how modern language models work and will be able to use the main interaction techniques for generative AI systems effectively and responsibly.
## Theoretical Knowledge
Students will understand:
- the fundamental principles underlying neural networks and Large Language Models;
- the role of the Transformer architecture in modern language processing systems;
- similarities and differences between the functioning of AI systems and certain human cognitive and linguistic processes;
- the main prompt engineering techniques and the criteria for selecting the most appropriate techniques for different contexts;
- the main limitations of language models, including bias, hallucinations, and vulnerability to prompt injection;
- the main applications of generative AI in society and the workplace, as well as their economic and social impacts;
- the main ethical issues associated with the use of AI and the principles for its responsible use.
### Practical Skills
Students will be able to:
- design prompts appropriate to specific objectives and contexts of use;
- apply prompt engineering techniques to improve the quality, reliability, and controllability of model outputs;
- identify the main limitations of language models and adopt strategies to mitigate them;
- create customized conversational assistants;
- use language models within simple workflows and pipelines to automate language-based tasks;
- create digital artifacts and small-scale projects using generative AI tools.
No prior programming knowledge is required.
During the course, students will carry out practical exercises and project-based activities in class. These activities will contribute to the overall assessment in preparation for the final written examination.
Programme
The laboratory aims to provide students with a solid theoretical foundation in Artificial Intelligence applied to language, together with advanced skills in interacting with Large Language Models (LLMs) through the main prompt engineering techniques, the creation of customized conversational assistants, and the development of simple agentic systems.These objectives will be pursued through two complementary phases.
The first phase, mainly theoretical, will focus on understanding how neural networks work and, in particular, on the Transformer architecture underlying modern Large Language Models. Students will examine the fundamental principles that enable these systems to process and generate natural language, as well as their main limitations and risks.
Topics covered will include, among others, bias, hallucinations, prompt injection, human-machine interaction, human-in-the-loop approaches, and the role of language in interaction with Artificial Intelligence systems. The social, economic, and ethical impacts arising from the widespread adoption of generative AI will also be discussed.
The second phase, mainly practical, will focus on prompt engineering techniques and their application across the different use cases presented during the course. Students will directly experiment with language models and use these tools to carry out exercises and create digital artifacts, customized assistants, and simple AI-based workflows.
# Learning Objectives
By the end of the laboratory, students will have acquired both a theoretical and practical understanding of how modern language models work and will be able to use the main interaction techniques for generative AI systems effectively and responsibly.
## Theoretical Knowledge
Students will understand:
- the fundamental principles underlying neural networks and Large Language Models;
- the role of the Transformer architecture in modern language processing systems;
- similarities and differences between the functioning of AI systems and certain human cognitive and linguistic processes;
- the main prompt engineering techniques and the criteria for selecting the most appropriate techniques for different contexts;
- the main limitations of language models, including bias, hallucinations, and vulnerability to prompt injection;
- the main applications of generative AI in society and the workplace, as well as their economic and social impacts;
- the main ethical issues associated with the use of AI and the principles for its responsible use.
### Practical Skills
Students will be able to:
- design prompts appropriate to specific objectives and contexts of use;
- apply prompt engineering techniques to improve the quality, reliability, and controllability of model outputs;
- identify the main limitations of language models and adopt strategies to mitigate them;
- create customized conversational assistants;
- use language models within simple workflows and pipelines to automate language-based tasks;
- create digital artifacts and small-scale projects using generative AI tools.
No prior programming knowledge is required.
During the course, students will carry out practical exercises and project-based activities in class. These activities will contribute to the overall assessment in preparation for the final written examination.
Core Documentation
They will be announced soon