ML — LLMpedia

ML
Name	Machine learning
Field	Artificial intelligence, Computer science, Statistics
Invented	1950s
Notable inventors	Arthur Samuel (computer scientist), Tom M. Mitchell, Geoffrey Hinton, Yann LeCun, Yoshua Bengio
Institutions	Massachusetts Institute of Technology, Stanford University, University of Toronto, Carnegie Mellon University, Google

Contents

ML is the study and engineering of algorithms that improve performance on tasks through data-driven adjustments rather than explicit programmatic rules. It overlaps with Artificial intelligence, Statistics, Data science, and Signal processing, and it underpins technologies deployed by Google, Microsoft, Amazon (company), Apple Inc., and OpenAI. Practitioners draw on methods from Neural networks, Bayesian statistics, Optimization (mathematics), and Information theory to build systems used across Silicon Valley firms, research labs at IBM, and government research agencies like DARPA.

Definition and scope

The field includes supervised paradigms exemplified by tasks studied at conferences such as NeurIPS, ICML, CVPR, and ACL; unsupervised paradigms linked to work by researchers at Bell Labs, Bell Labs Research, and AT&T; and reinforcement approaches advanced by teams at DeepMind and OpenAI. Technical scope spans models like Decision tree, Support vector machine, Hidden Markov model, Convolutional neural network, and Transformer (machine learning model), and tools from libraries such as TensorFlow, PyTorch, scikit-learn, and JAX. Applications integrate with systems deployed by Tesla, Inc., Uber, Airbnb, Salesforce, and Facebook.

Origins trace to early work by Alan Turing and implementation efforts by Arthur Samuel (computer scientist) at IBM, with theoretical foundations from Donald Hebb and algorithms refined in the 1960s and 1970s at institutions like MIT and Stanford University. The 1980s saw revival via backpropagation popularized by researchers at Bell Labs and University of Toronto, later accelerated by breakthroughs from Geoffrey Hinton, Yann LeCun, and Yoshua Bengio in the 1990s and 2000s. The 2010s deep learning surge coincided with large-scale experiments by Google DeepMind, industry projects at Facebook AI Research, and datasets such as ImageNet fueling progress noted at NeurIPS and ICML. Contemporary development includes contributions from international labs like Microsoft Research, Amazon Web Services, Alibaba Group, and government-funded centers in China, United Kingdom, and Canada.

Core supervised techniques include regression methods advanced by statisticians at University of Chicago and classification algorithms used in systems deployed by Adobe Systems and LinkedIn. Ensemble methods such as boosting and bagging were developed in research by teams associated with University of California, Berkeley and practitioners at H2O.ai. Unsupervised learning encompasses clustering approaches from scholars at University of Cambridge and dimensionality reduction methods like Principal component analysis applied in projects at Siemens. Probabilistic methods draw from work at Columbia University and New York University, while reinforcement learning traces to classic control theory labs at MIT and operational deployments by NVIDIA. Deep architectures—recurrent models, convolutional stacks, and attention-based transformers—were popularized through collaborative research at Stanford University, University of Toronto, Carnegie Mellon University, and industrial labs including Google Research.

Technique transfer has yielded systems in healthcare used by startups collaborating with Mayo Clinic and Johns Hopkins University, finance products deployed by Goldman Sachs and JPMorgan Chase, and recommendation engines powering services at Netflix, Spotify, and YouTube. Autonomous vehicle research is led by teams at Waymo, Cruise (company), and Tesla, Inc., integrating perception modules developed with tools from NVIDIA and mapping efforts by HERE Technologies. Natural language systems stem from work at OpenAI, Google, and Microsoft Research and are applied in products by Salesforce, Zoom, and Adobe Systems. Scientific discovery efforts include collaborations with National Institutes of Health, CERN, and NASA.

Concerns address bias exposed in deployments at companies like Amazon (company) and regulatory scrutiny in jurisdictions including the European Union and United States Congress. Privacy debates involve frameworks promulgated by General Data Protection Regulation authorities and legal cases heard in courts in California and United Kingdom. Safety and alignment research is undertaken at organizations such as OpenAI, DeepMind, Future of Humanity Institute, and policy centers at Harvard University and Stanford University. Workforce impacts and education initiatives link to programs at Coursera, edX, and university departments in China and India. Public discourse often references reports from UNESCO, OECD, and national science agencies debating standards, accountability, and governance.