As part of the University of Josip Juraj Strossmayer Osijek Career Week, the Center for Artificial Intelligence of FERIT is organizing an AI meet.ing on May 9, 2024.

• Time: 10:00 -12:00,
• Duration: 90 minutes
  (three lectures of 20 minutes each + questions from the audience),
• Room: K2-12,

At the AI meet.ing, three lectures by FERIT scientists from different areas of artificial intelligence will be held.
All students, employees and other audiences interested in artificial intelligence are invited.

Program:

• prof. Marijan Herceg, PhD – The main challenges in autonomous driving (duration 20 min)

Autonomous vehicles represent a revolutionary technological innovation that will have a significant impact on life and the economy in the future. However, to successfully integrate this technology, several complex challenges need to be solved. Key challenges include ensuring safety, which requires the development of reliable sensors and artificial intelligence algorithms that are adapted to changing road conditions. Clear legal regulations must also be created to ensure the responsible use of autonomous vehicles, taking ethical and moral aspects into account. Furthermore, it is necessary to provide infrastructures such as smart roads to enable the smooth functioning of autonomous vehicles. And finally, to ensure economic sustainability, business models must be found that enable the mass introduction of this technology.

• prof. Robert Cupec, PhD – Training neural networks on automatically generated data sets (duration 20 min)

Learning neural networks for machine vision usually requires a large data set, which is usually labeled manually. Manual labeling takes a lot of time and is subject to human error. For this reason, the automatic data labeling for neural network learning is an extremely interesting possibility which is the subject of numerous scientific studies. This lecture will present several examples of data sets automatic generation for learning neural networks for application in the realization of machine vision for service robots and analyze the influence of various factors in the automatic generation of synthetic 3D images on the accuracy of object recognition and classification.

• Assistant prof. Hrvoje Leventić, PhD – Prompt is all you need (duration 20 min)

In today’s world, where large-scale language models (LLMs) have become key tools in natural language processing and artificial intelligence, the importance of precisely formulated queries has never been greater. The lecture “Prompt is all you need” takes you on a journey through advanced methods and strategies for querying large language models. The aim of the lecture is to equip listeners with the knowledge and skills necessary to achieve better quality responses from LLMs, exploring how to properly construct prompts that will encourage models to provide precise, relevant and useful information. Using examples, we will show how prompt fine-tuning can affect model responses, and how to use specific characteristics and capabilities of LLMs to solve complex tasks.