The diverse group of partners working together on THEMIS 5.0 is a part of what makes this project so important. 'Get to Know THEMIS 5.0' is a campaign designed to communicate the core values and aims of the project, through the lens of the unique stories and experiences of the people who make THEMIS possible. By hearing their stories, we hope that readers can understand the importance of improving trust in AI not just for themselves, but for citizens across the world.
What is your name, where do you work, and what is your role in THEMIS 5.0?
Dimitris Apostolou, Senior Research Scientist at ICCS.
I will be leading WP2 which aims to understand the socio-economic and human factors that determine trustworthiness of hybrid decision support in socio-technical systems; explore the AI SoTA relevant to the trustworthiness evaluation and optimisation; establish the legal and ethical framework to an AI-driven trustworthiness optimisation; determine the user requirements for a human-centred trustworthiness optimisation environment; and specify the related decision-making processes for the development of the needed AI models.
Why is the THEMIS 5.0 project important to you?
The THEMIS project holds significant importance for my research, which involved the design, development, and evaluation of a variety of AI methods, tools, and algorithms with an emphasis on decision making Leveraging the THEMIS developed methodologies for assessing the human footprint of AI can guide my research towards ethical and responsible AI development.
Have there been any situations (work or personal) where you have had concerns about bias or fairness in an AI system or outcome?
Some concerns arise in the field of discrimination regarding AI systems. Specifically AI algorithms used in recruiting and hiring systems. These algorithms could be biased against certain groups or demographics because they learn from historical data in the related field. Another implication could occur when these algorithms are used in loan approval and credit score.
In your opinion, what are the key challenges to enhancing trustworthiness in AI systems?
Bias and fairness are a key challenge since AI algorithms are trained with historical data that include biases against certain groups and demographics. Another key challenge is maintaining privacy in AI systems since they handle a lot of sensitive data (medical, financial etc.). Thus the robustness and the resilience of AI systems against malicious attacks needs to be concerned.
How do you stay updated on the latest advancements and practices in AI?
In order to stay updated on the latest advancement and practices in AI as a research scientist I prioritize to keep track of new academic publications and research papers and whenever is possible contribute myself to open projects. I also think it’s very important to attend related conferences and workshops to be in touch with the community.
If you could use AI to do something in your everyday life (which it doesn’t already do!) what would it be?
If given the opportunity, I would love to utilize AI to enhance our education activities. Specifically, I envision an AI-powered personal assistant that not only manages my lecture schedule and lab tutoring but also assists in the preparation and delivery of courses. This AI assistant could analyse my daily routines, tasks, and obligations to provide tailored recommendations and reminders, helping me optimize my time. Additionally, it could assist in automating repetitive tasks, such as scheduling appointments with students, responding to emails, or managing departmental duties, allowing me to focus on more meaningful and creative pursuits. AI could also scan our library and the Web for emerging new educational content that I use utilise readily in my courses, such as new assignments, quizzes, or lab exercises.
Commentaires