Far off in an unfashionable municipality of the galaxy lies an insignificant blue-yellow planet whose eight billion life forms are so rudimentary that they still think they are brilliant beings, and their digital twins might be a splendid idea ^{[cf. 5]}. Several twins would enable them to do several tasks at once. They could spend their entire day with meeting friends and relaxing. The research field associated with this idea is called artificial intelligence (AI), i.e., "the theory and development of computer systems able to perform tasks that normally require human intelligence [...]" ^[16].

Tasked with raising curiosity and enthusiasm, Professor Gerard de Melo as the chair of Artificial Intelligence and Intelligent Systems at Hasso Plattner Institut (HPI) was entrusted with introducing students to the recent trends in AI. Topics of his chair include Natural Language Processing, Information Extraction and Information Retrieval, Knowledge and Data Resources, and Multimodality ^[4]: One of his research goals addressed the concept of explainability in AI using knowledge graphs ^[13].

The following paragraph will be dedicated to an educational introduction about AI of the local life forms history. For those who are already AI-lien experts, please feel free to skip this part to learn more about a limited selection of the chairs research and the planet's current trends.

Emergence and terrestrial classification of AI

"You have to know the past to know the present" ^[24] and predict future trends: The birthplace of AI was in the Golden Age 1940-1955, while the foundation for this work can be traced back to espionage and deciphering codes during the second world war 1939-1945. Alan Turing is known for contributing to crack the Enigma device code ^[31][21]. Warren McCulloch and Walter Pitts synthesized the logical mechanism of human intelligence, the biological neuron ^[19]. Based on previous research, Turing and John von Neumann proposed the architecture of contemporary machines ^[2]. Claude Shannon's (theoretical) paper about a computer to play chess inspired ^[28] Arthur Samuel to develop a scoring system which used the min-max strategy for the chances of someone winning in a game of draughts, a project for IBM (International Business Machines Corporation) ^[25]. Arthur Samuel coined the phrase 'machine learning' (ML) in 1952, defining it as the "ability of a computer to perform a task itself without being explicitly programmed" ^[26]. ML later became a sub-field in AI research. In the summer of 1956, some research fellows, John McCarthy, Marvin Minsky, Nathaniel Rochester and Claude Shannon, organized the "Dartmouth Conference" at New Hampshire; during that Conference, McCarthy was the first to mention the term artificial intelligence ^[21][20]. Even in the outer space galactic museum, Frank Rosenblatt's building 1957, a Mark 1 (single-layer) perceptron can be appraised and most have given up untangling the machine. To achieve this, Rosenblatt combined Donald Hebb's model of brain cell interaction with Arthur Samuel's machine learning efforts ^[23]. The amount of terrestrial knowledge processable was due to the human's limited GPU power, insufficient digital data and unpolished concepts - all not enough to even differentiate between their own facial patterns ^[21]. In 1964, Joseph Weizenbaum developed the first chatbot, Eliza, at the MIT (Massachusetts Institute of Technology) based on rule-based systems, which are a set of if-else rules combined with symbolic logic ^[33][17]. Implementing the machine learning concept in 1997, IBM's Deep Blue defeated Garry Kasparov, the world's reigning chess champion at that time ^[14]. Simple rule-based systems had their limitations for image pattern recognition ^[10], e.g., recognizing animals like the Tribleustes ventricosus on pictures ^[22]. 'Deep learning' allows "computers to learn from experiences and understand the world in terms of a hierarchy of concepts, by building them out of simpler ones. If we build a graph showing how these concepts are build on top of each other, the graph is deep" ^[10] which explains the name ^[10]. Geoffrey Hinton's research group has engineered a deep convolutional neural network (CNN) outperforming the competition in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) ^[12]. The innovation's in AI revolutionized the interaction with technology and interactions within the planet. The Alan Turning Institute founded in 2015 in London, added AI to their remit in 2017 ^[29]. Same year, in memory of Weizenbaum's achievements, the German federal ministry of education and research founded the Weizenbaum Institute in Berlin ^[32]. There are currently four KI service centers within Germany and HPI is one of these ^[7]. Internationally, initiatives like the EU AI Act aim to regulate AI with respect to safety, transparency, and accountability, by categorizing AI systems based on risk, setting stringent standards. Within this legislative framework, explainable AI plays a crucial role in ensuring that AI decisions are transparent and understandable. This transparency is essential for building trust in AI outcomes, and enhancing accountability in compliance with regulatory standards ^[1][27].

Figure 1: Timeline of the AI journey ^[11][30]
(Data from Heutger & Kückelhaus (2018) original data from Lavenda, D. (2017) / Marsden, P. (2015) no longer available c. Toosi et al. (2021))

Curiosity driven research projects

AI cannot only improve efficiency, but it might also expand the perception of the world and create new content: A use case where this is especially important is for people with disabilities, such as visual impairment. Prof. de Melo's research focuses on developing software that allows to answer specific questions on the basis of any kind of pictures taken. Providing information about pictures and video material has also been addressed by a Danish start up company, called ‘bemyeyes’, in 2015. They are currently beta testing ‘bemyAI’ in a cooperation with OpenAI Chat-GPT4 which you might want to keep an eye out ^[3], and see if some of their research ideas will be implemented. Expanding mankind's perception of the world could help on the roadmap ahead. Using video material to kinesiological movement patterns could for example be used to identify cultural differences, recognition perceptions and reactions to specific movement, or even warn about potentially dangerous situations. It might even be possible to use audiovisual content to warn someone about specific situations and give instructions in emergency situations.

A critically endangered species on the planet are the Gorillas (IUCN List). Terrestrial scientists have been going bananas as this species achieved eudaimonia without using terrestrial life's technology, eating bananas and living in the jungle for millenniums. The 'Savina Plattner Ethical Charities' (SPAC) and Magdalena Bermejo work very hard to track the traces of the gorillas. Magdalena Bermejo lived for a decade in the rainforest to track gorillas and got acquainted with all gorillas within this rainforest and learned some of their secret arts. To prevent the knowledge of Magdalena Bermejo getting lost, Prof. de Melo worked together with undergraduate students to retain her knowledge. They teamed up with the 'ConservationXLabs' creating low powered devices that collect video information. They use a dataset with the Gorillas' names from Megdalena Bermejo, who is an expert capable of differentiating all gorillas by names. Using image recognition the AI can learn to differentiate between them as well, such that Megdalena's ability becomes widely available for society.

Those who are interested to learn more about AI research should take the opportunity to visit Prof. de Melo on campus. For instance, there is a growing concern about fairness in intelligent decision-making systems, particularly in recommendation systems, where biases and disparities in performance can occur. Prof. Gerard de Melo and colleagues address these challenges in their research by proposing a fairness-constrained heuristic re-ranking approach for explainable recommendation systems based on knowledge graphs ^[8]. Their approach, such as the Graph-Based Decomposition method, employs fairness-aware algorithms to enrich recommendation diversity and accuracy. This method not only enhances transparency in AI recommendations but also mitigates biases that can arise from limited training data and user activity preferences (Fig2).

Figure 2: Case study of recommendation paths, before and after adding the fairness algorithm^[8].

So, dear AI-liens, as you wander this energy robbing and perplexing roadmap of AI-Trends, remember that the pathway is less about your final destination and more about the your enjoyable and wonderful journey. These include clunky algorithms that could not distinguish a tribbles ^[22] from a toaster, to the dizzying heights of neural networks capable of creating artistic masterpieces, debating politics and extending our perspective closer to the mysterious concept of 42 ^{[cf. 5]}. Overall, Prof. Gerard de Melo's and his colleagues' study demonstrates that integrating fairness constraints into AI systems not only improves recommendation quality and transparency but also reduces unfairness in decision outcomes, thereby fostering greater trust and satisfaction among users. There has been an ongoing debate whether AI on planet Earth would help to foster Diversity, Equity, and Inclusion (DEI) or lead to further divergence, which is heavily affected by the design and implementation of these systems. Which will be crucial for further implementation like AI politicians ^[6], deep patients' digital twins ^[15].

Thus, as you embark on your very own professional quest, don’t panic with your digital towels, and always be ready to hitch a ride on upcoming big innovation. AI definitely is similar to the universe expanding in ways we can scarcely imagine ^{[cf. 5]}., so stay tuned for parallel emerging AI-trends, such as Quantum Machine Learning (QML) ^[18] or Neurosymbolic AI ^[9]. After all enjoy your Friday's beer with your friends and colleagues in the grand cosmic scheme and do not send your digital twins to these intriguing courses. Cheers to an AI Summer!

References

^[1] Almada, M. & Petit, N. (2023). The EU AI Act: a medley of product safety and fundamental rights? Robert Schuman Centre for Advanced Studies Research Paper No. 2023/59. ssrn.com/abstract=4308072 or dx.doi.org/10.2139/ssrn.4308072

^[2] Al-Hashimi, H. M. (2023). Turing, von Neumann, and the computational architecture of biological machines. Proc Natl Acad Sci USA, 120(25), e2220022120. doi.org/10.1073/pnas.2220022120

^[3] Be my eyes (n.d.). Be my eyes.https://www.bemyeyes.com/blog/introducing-be-my-eyes-virtual-volunteer .2024-07-05.

^[4] de Melo, G. (2022). Chair of Artificial Intelligence and Intelligent Systems. gerard.demelo.org .2024-07-05.

^[5] Douglas, A. (1980). The Hitchhiker's Guide to the Galaxy. Harmony Books. New York.

^[6] Endacott, S. (2024). Vote ai steve - independent ai candidate for brighton. www.ai-steve.co.uk .2024-07-05.

^[7] Federal Ministry of Education and Research (2022). Foerderung von vier KI-Servicezentren gestartet. www.bmbf.de/bmbf/shareddocs/kurzmeldungen/de/2022/11/foerderung-von-4-ki-zentren-gestartet.html .2024-07-05.

^[8] Fu, Z., Xian, Y., Gao, R., Zhao, J., Huang, Q., Ge, Y., Xu, S., Geng, S., Shah, C., Zhang, Y., & De Melo, G. (2020). Fairness-Aware Explainable Recommendation over Knowledge Graphs. In SIGIR 2020 - Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval (pp. 69-78). (SIGIR 2020 - Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval). Association for Computing Machinery, Inc. doi.org/10.1145/3397271.3401051

^[9] Garcez, A. & Lamb, L. (2023). Neurosymbolic AI: the 3rd wave. Artificial Intelligence Review. 56. 1-20. 10.1007/s10462-023-10448-w.

^[10] Goodfellow, I., Bengio, Y. and Courville, A. (2016) Deep Learning. Adaptive computation and machine learning. MIT Press. London.

^[11] Heutger, M., & Kückelhaus, M. (2018). Artificial Intelligence in Logistics: A Collaborative Report by DHL and IBM on Implications and Use Cases for the Logistics Industry. DHL Customer Solutions & Innovation.

^[12] Hinton, G. E. & Salakhutdinov, R. R. (2006). Reducing the Dimensionality of Data with Neural Networks.Science,313,504-507.DOI:10.1126/science.1127647

^[13] Hogan, A., Blomqvist, E., Cochez, M., D’amato, C., De Melo, G., Gutierrez, C., Kirrane, S., Gayo, J. E. L., Navigli, R., Neumaier, S., Ngomo, A. N., Polleres, A., Rashid, S. M., Rula, A., Schmelzeisen, L., Sequeda, J., Staab, S., & Zimmermann, A. (2021). Knowledge Graphs. ACM Computing Surveys, 54(4), 1–37. doi.org/10.1145/3447772

^[14] Kasparov, G. (2018). Deep thinking: Where Machine Intelligence Ends and Human Creativity Begins. John Murray.

^[15] Katsoulakis, E., Wang, Q., Wu, H., Shahriyari, L., Fletcher, R., Liu, J., Achenie, L., Liu, H., Jackson, P., Xiao, Y., Syeda-Mahmood, T., Tuli, R., & Deng, J. (2024). Digital twins for health: a scoping review. NPJ digital medicine, 7(1), 77. doi.org/10.1038/s41746-024-01073-0

^[16] Knowles, E. M. (2005b). Oxford Dictionary of Phrase and Fable. Oxford University Press, USA.

^[17] MacColl, H. (1905). VI.—SYMBOLIC REASONING (VI.)1. Mind, XIV(1), 74–81. doi.org/10.1093/mind/xiv.1.74

^[18] Martín-Guerrero, J. D., & Lamata, L. (2022). Quantum Machine Learning: A tutorial. Neurocomputing, 470, 457–461. doi.org/10.1016/j.neucom .2021.02.102

^[19] McCulloch, W. S., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. the Bulletin of Mathematical Biophysics, 5(4), 115–133. doi.org/10.1007/bf02478259

^[20] Moor, J. (2006). The dartmouth college artificial intelligence conference: The next fifty years. AI Magazine, 27(4),87–91.

^[21] Muthukrishnan N, Maleki F, Ovens K, Reinhold C, Forghani B, Forghani R. Brief History of Artificial Intelligence. Neuroimaging Clin N Am. 2020 Nov;30(4):393-399. doi: 10.1016/j.nic.2020.07.004

^[22] Pevney, J. (director) (1967). The trouble with Tribbles.Desilu Productions. Paramount Studios.

^[23] Rosenblatt, F. (1961). Principles of Neurodynamics: Perceptrons and the Theory of Brain Mechanisms. Cornell Aeronautical Laboratory, Inc. Buffalo NY.

^[24] Sagan, C. E. (1980). Episode 2: One voice in the cosmic fugue. In Adrian Malone(Producer), Cosmos: A Personal Voyage, 1980. Television series episode

^[25] Samuel, A. L. (1959). Some studies in machine learning using the game of checkers. IBM Journal of Research and Development, 3(3), 210–229. doi.org/10.1147/rd.33.0210

^[26] Samuel, A. L. (1967). Some Studies in Machine Learning Using the Game of Checkers. II—Recent Progress. IBM J. Res. Dev., 44, 206-227.

^[27] Sánchez, P. M., Huertas Celfrán, A., Martínez Pérez, E.T., Demeter, D., Bovet, G., Martínez Pérez, G., & Stiller, B. (2022). Analyzing the robustness of decentralized horizontal and vertical federated learning architectures in a non-IID scenario. Appl Intell 54, 6637–6653. doi.org/10.1007/s10489-024-05510-1

^[28] Shannon, C. E. (1950). Programming a computer for playing chess. Philosophical Magazine, 41(4), 256–275.

^[29] The Alan Turing Institute (2024). We are the UK’s national institute for data science and artificial intelligence. www.turing.ac.uk. 2024-07-05.

^[30] Toosi, A., Bottino, A. G., Saboury, B., Siegel, E., & Rahmim, A. (2021). A brief history of AI: How to Prevent Another Winter (A critical Review). PET Clinics, 16(4), 449–469. doi.org/10.1016/j.cpet.2021.07.001

^[31] Tyldum, M. (2014) The imitation game.The Weinstein Company.

^[32] Weizenbaum Institut (n.d.). Weizenbaum institut.https://www.weizenbaum-institut.de/.

^[33] Weizenbaum, J. (1966). ELIZA—a computer program for the study of natural language communication between man and machine. Communications of the ACM, 9(1), 36–45. doi.org/10.1145/365153.365168