DIKU Alumni Day 2023: AI, ML and ChatGPT – risks and potential
DIKU Alumni Day focuses on Generative AI and ChatGPT
DIKU Alumni Association invites you to the annual Alumni Day in collaboration with DIKU Business Club. This year's theme will be 'AI, ML and ChatGPT – risks and potential '.
Again this year, you can ask to sit next to people from your own year of enrolment. Sign up by clicking on the red box located on the right side of this page. Last chance to sign up is Sunday 24 September.
Program:
| 17:00-17:10 |
Welcome by the Head of Department
What's happening at DIKU?
|
|
| 17:10-17:30 |
ChatGPT in Education
Professor Morten Misfeldt and Associate Professor Daniel Spikol will present interesting insights about ChatGPT and the entry of 'generative AI' into the education sector – what are the opportunities and what are the pitfalls? In the talk we will look at the challenges that Generative AI has been posing to formal education over the last years time and how these challenges resembles previous introduction digital tools in education. After that we will talk about the possibilities and threats that these tools poses to learning and life. |
Associate Professor of Computational Thinking at the Center for Digital Education, Department of Computer Science and the Department of Science Education |
| 17:30-17:50 |
What can we do about AI?Generative artificial intelligence supercharges existing concerns about the role of digital technologies in society, and forces us to engage with pressing questions about ethics, responsibility and values. This talk highlights the challenges and choices that we face as universities and societies in times of artificial intelligence.
|
|
|
Short break
|
||
| 18:00-18:15 |
The End of Words? Language Modelling from Pixels Language models are defined over a finite set of inputs, which creates a bottleneck if we attempt to scale the number of supported languages. Tackling this bottleneck usually results in a trade-off between what can be represented in the embedding matrix and computational issues in the output layer. I will present the Pixel-based Encoder of Language, which does not suffer from this issue, making it possible to directly support any language that can be typeset by a computer. |
 Desmond Eliott, Tenure-track Assistant Professor in the Department of Computer Science, Natural Language Processing |
| 18:15-18:30 |
AI, ML and Outliers - A Case Study of Asylum Case Data in Denmark
Refugees worldwide now face an evolving landscape of data-driven decision-making as they seek asylum. Countries like the US, Canada, and Australia are increasingly turning to machine learning algorithms to predict asylum outcomes, aiming to mitigate potential biases among human judges and achieve greater decision consistency. However, these algorithmic approaches often hinge on notions of "norms" and rely on assumptions of representativeness, which can profoundly affect the lives of vulnerable individuals. This presentation takes a closer look at the intricate concept of outliers within this context and delves into how machine learning algorithms, in conjunction with human judgment, construct the ideas of "normal" and "outlier."
Through illustrative examples of asylum case outliers, we emphasize that amidst the fervor of big data, the data points representing real people can be easily overlooked when they don't align with the algorithmically established norm. Our objective is to illuminate the complexity of asylum decision-making, shedding light on potential blind spots that may arise when employing AI in high-stakes areas like this one.
|
Kristin Kalten-häuser,
PhD Student, Computer Science, Software, Data, People and Society
|
| 18:30-18:45 |
Panel discussion and Q&A
|
|
| 18:45-19:00 |
DIKU revue 50th anniversary sketch
The DIKU-Revue serves up some revue classics as a warm-up for their upcoming 50th anniversary, which will be celebrated in October. For example, you can hear: Torben and Torben performing "A workplace in Denmark"
|
|
| 19:00 - | Dinner |
Jakob Grue Simonsen, Department chair and Professor at DIKU
Professor Morten Misfeldt, r
