Generative Artificial Intelligence

Recommended Prerequisites

Machine Learning, Deep and Reinforcement Learning, Python programming; Good English reading, writing and speaking skills; Mathematics and statistics.

Teaching Methods

During the lectures (T) the concepts, the theories, the algorithms will be presented and discussed. In the lab classes (PL) students will consolidate what was learned in T. The practical assignments will be done under the supervision of the teacher. Grading will be based on two components: (1) projects involving the techniques; (2) a research work.

Learning Outcomes

The UC aims to study and develop generative AI models using connectionist, evolutionary, and biological approaches. Based on previous knowledge acquired throughout the course, the main challenges and opportunities in this field will be analyzed. Emphasis will be given to hands-on learning, following a Project Based Learning approach.

By the end of the unit, the student will have a comprehensive understanding of the generative AI field and will also be capable of developing and/or adapting generative systems to address real-world needs, applying these approaches in various domains.

The main competencies to be developed are:

Instrumental – analysis and synthesis, problem-solving

Personal – critical thinking

Systemic - practical application of the theoretical knowledge; research

The secondary competencies are:

Instrumental – organizing and planning

Personal – work in teams

Systemic – autonomous learning; creativity.

Work Placement(s)

No

Syllabus

1. Introduction to Generative AI
- Overview and applications
- Distinction between generative and discriminative models
- Historical context and evolution
2. Classical Generative AI Methods
- Rule-based and expert systems
- Production systems
- Constraint logic programming
- Evolutionary
- Markov chains (for generativity)
3. Autoencoders
- Autoencoders the basics
- Non-linear dimensionality reduction
- Variational Autoencoders
4. Generative Adversarial Networks
5. Diffusion Models
6. Generative ML Exploration Techniques
- Latent space exploration
- Prompt generation and improvement
- Optimisers
- Quality-Diversity
7. Text generation and large language models
8. Generative AI State of the Art
- Image generation and synthesis
- Music and sound generation
- Video generation and synthesis
- Multimedia and hybrid models
9. Ethical Considerations, Challenges, Opportunities.

Head Lecturer(s)

Fernando Jorge Penousal Martins Machado

Assessment Methods

Assessment
Research work: 30.0%
Project: 70.0%

Bibliography

Foster, D. (2023). Generative Deep Learning (2nd ed.). O’Reilly Media. ISBN: 97810981341811

Bentley, P. J., & Corne, D. W. (Eds.). (2002). Creative Evolutionary Systems. Morgan Kaufmann Publishers

Bentley, P. J. (Ed.). (1999). Evolutionary Design by Computers. Morgan Kaufmann Publishers

Cope, D. (2004). Virtual Music: Computer Synthesis of Musical Style. The MIT Press. ISBN: 97802622559431