Definition
AGI is defined as a machine intelligence that is capable of understanding and learning any intellectual task that can be performed by a human being. In contrast to weak AI, AGI strives for a comprehensive, flexible intelligence that enables it to think, learn and be creative independently (Bostrom, 2014).
Historical development
The idea of artificial intelligence dates back to antiquity, as the myth of Talos shows. However, the modern era of AI research did not begin until the middle of the 20th century. A decisive moment was the Dartmouth Conference in 1956, which is often referred to as the birth of AI research (McCarthy et al., 2006).
Technological foundations and challenges
Machine learning and deep learning are at the heart of modern AI development. Despite impressive progress in some areas, fundamental limitations remain. As the lack of generalisation ability contrasts with the human ability to seamlessly apply what has been learned to novel contexts (Lake, Salakhutdinov, & Tenenbaum, 2017).
New research approaches
Innovative approaches are being researched to overcome these challenges. These include differentiable neural computers and probabilistic programming (Ha et al., 2016). These could lead to a more causal and robust modelling of knowledge than the purely correlation-based approach of current deep learning systems (Pearl, 2009).
Global research landscape
AGI research is a global endeavour. Leading players such as OpenAI, DeepMind and Meta are driving development forward. OpenAI, for example, is working to develop guidelines and standards to ensure that AGI technologies are used in the best interests of humanity (OpenAI, 2021).
Ethical challenges and governance
The development of AGI raises important ethical questions. A key concern is that AGI systems that outperform human intelligence could get out of control and make decisions (singularity) that are not in line with human values and interests (Bostrom, 2014). To address this, various governance models are being discussed, including the development of a ‘constitutional AI’ and international regulatory frameworks such as the EU AI Act (European Commission, 2021).
Social impact
The potential impact of AGI on society is far-reaching. AGI will lead to a massive restructuring of the labour market and personalised learning by AGI tutors will revolutionise the educational landscape (Brynjolfsson & McAfee, 2014). Moreover, AGI will fundamentally challenge our understanding of intelligence, consciousness and the nature of being human’ (Chalmers, 1996).
References
Bostrom, N. (2014). Superintelligence: Paths, Dangers, Strategies. Oxford University Press.
Brynjolfsson, E., & McAfee, A. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. W. W. Norton & Company.
Chalmers, D. J. (1996). The Conscious Mind: In Search of a Fundamental Theory. Oxford University Press.
European Commission. (2021). Proposal for a Regulation of the European Parliament and of the Council Laying Down Harmonized Rules on Artificial Intelligence (Artificial Intelligence Act). Brussels.
Ha, D., Dai, A. M., & Le, Q. V. (2016). HyperNetworks. In Proceedings of the International Conference on Learning Representations.
Lake, B. M., Salakhutdinov, R., & Tenenbaum, J. B. (2017). Human-level concept learning through probabilistic program induction. Science, 358(6365), 1046-1050.
McCarthy, J., Minsky, M., Rochester, N., & Shannon, C. (2006). A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence. In AI Magazine.
OpenAI. (2021). OpenAI API. Retrieved from https://openai.com/api.