The Alignment Problem: Machine Learning and Human Values

Tim Brennan is a Scottish statistician and criminologist who, along with his colleague Dave Wells, co-founded Northpointe, an organization known for creating the Correctional Offender Management Profiling for Alternative Sanctions (COMPAS). This algorithmic tool is used widely throughout the United States to assess the likelihood of a defendant reoffending. Brennan’s extensive background in psychology and statistics equipped him with the knowledge to develop a risk assessment tool that analyzes various behavioral and psychological aspects. His work on the COMPAS tool positions him as a significant figure in the intersection of technology, criminal justice, and ethics, especially in discussions around the use of algorithmic decision-making in the justice system. The widespread implementation of COMPAS has sparked significant debate regarding the fairness, transparency, and accuracy of using algorithmic methods in criminal sentencing and parole decisions, highlighting the Ethical Implications of AI Use.

In The Alignment Problem, the work of Tim Brennan and Northpointe’s COMPAS tool represents a practical example of the challenges and ethical dilemmas inherent in the use of AI and machine learning within critical societal frameworks. In Chapter 1, Christian recounts the dispute between Northpointe and the investigative journalism organization ProPublica regarding the issues of bias and the opacity of the algorithms that ProPublica’s research unveiled.

Alex Krizhevsky is a Ukrainian-Canadian computer scientist known for his work in the field of deep learning and AI. Krizhevsky earned his PhD in computer science at the University of Toronto, where he studied under the mentorship of Geoffrey Hinton. While still a graduate student, he co-developed with his colleague Ilya Sutskever a deep convolutional neural network that significantly advanced the capabilities of AI in image recognition. This network, known as AlexNet, marked a breakthrough in the 2012 ImageNet competition, winning by a large margin over traditional computer vision methods.

AlexNet’s success demonstrated the power of deep learning models and has been credited with sparking the current interest and investment in deep learning technologies across various industries. In The Alignment Problem, Alex Krizhevsky’s work with AlexNet contributes to the book’s exploration of The Intersection of Human and Machine Learning by providing an example of how innovations in AI can lead to substantial shifts in the capabilities and applications of technology. Krizhevsky’s contribution illustrates the rapid progression of machine learning technologies and the accompanying challenges in ensuring these technologies align with human values and ethical standards. As AI systems like AlexNet become more potent and widespread, the alignment problem emphasizes the need to address the implications of such technologies, particularly in terms of transparency, fairness, and accountability.