Notebook

Appearance

Machine Learning ​

Machine Learning is the ability of a program or machine to think and learn by itself.

Artificial Intelligence ​

Machine Learning is used as a synonym of Artificial Intelligence, but it is actually a subset of the latter. AI define a broader scope, where systems can simulate human intelligence without relaying on Machine Learning (e.g. chat agents, video game bots, …).

The definition of AI is so large that we tend to classify it into sub-categories

TermAcronymApplication
Narrow AI-Limited to specific tasks (e.g. translation, facial recognition, etc.).
Artificial General IntelligenceAGIApplied to any problems, matching human cognitive capabilities
Artificial superintelligenceASICognitive applications superior to human being

Key events ​

  • 1949: Alan Turing already created the Turing test to evaluate a system's intelligence by having a user interact with it. If the user believes they are communicating with a human, the system passes the test.
  • 1955: Logic Theorist is consider the first AI program, build to perform automated reasoning.
  • 2016: AlphaGo defeat the best human player of Go in 2016
  • 2018: AlphaFold is a model that can predict the 3D structure of a protein from an amino acid sequence using reinforcement learning. In 4 years, it has published about 200 million proteins 3D structure predictions vs. 100’000 after 50 years of research using traditional methods.
  • The ELIZA effect describe how people tend to attribute human traits to intelligent systems that have a textual interface (typically with LLMs). Because of that, users empathise with AI system, from being polite to building friendship with.
  • The dead Internet theory is a conspiracy theory that state that the web (social media in particular) mainly consist of bots

Resources ​

Glossary ​

  • MLM – Machine Learning Model
  • ML Facilities – Resources used to train a model
  • ML Inference – Using a trained model on data
  • AI Glossary – NNGroup

Tools ​

Learn ​