AI Model Evaluation
Artificial intelligence (AI) model evaluation is the practice of measuring whether an AI model has a particular kind of ability or knowledge. The goals of evaluation range from assessing performance on a concrete task (e.g., spam email classification) to gathering evidence for a high-level cognitive
Algorithmic Bias
Algorithmic bias refers to prejudicial, discriminatory, unjust, inaccurate, or otherwise disparate performance or outcomes from algorithmic systems based on racial, gender, or other attributes of an individual or a group. The concept of algorithmic bias emerged at the intersection of computer scienc
Augmented Reality
Augmented reality (AR) is a technology that provides perceptual experiences that blend the user’s real-world context with computer-generated virtual elements to form a unified scene. AR experiences situate the virtual in relation to the real world and then dynamically track changes in perspective to
Computational Complexity
Designing effective computational systems is often a matter of finding ways in which simple logical operations can be combined to perform more complex tasks. Computer scientists therefore gauge the complexity of tasks by asking how many such operations would be needed to perform them. They are espec
Computational Models of Language Learning
A computational model of language learning is a formal description of how linguistic input can be transformed into either mental representations or (linguistic) behavior. For instance, a model may formally describe how a speech signal directed at a child can be processed to learn word-meaning mappin
Consciousness and AI
Consciousness is what distinguishes the familiar thoughts, emotions, and sensory experiences of waking life from the processes that go on in our brains without our awareness. Conscious experiences feel some way for the subject and seem to afford the subject a distinctive form of first-person access.
Large Language Models
A large language model (LLM) is a computational system, typically a deep neural network with a large number of tunable parameters (i.e., weights), that implements a mathematical function called a language model. A language model (LM), in its most general form, is a probability distribution over poss
Levels of Analysis
Complex information-processing systems can be understood or analyzed in terms of their function (what the system does and why), the processes and operations that achieve that function (how it does it), or the physical hardware that performs those processes and operations (how those processes and ope
Rational Analysis
Rational analysis is an approach to understanding human cognition that treats it as optimized to solve problems given certain constraints. For example, a rational analysis of memory starts from the idea that memory systems are optimized to the statistics of how the items to be remembered occur in th
Recurrent Neural Networks
Recurrent Neural Networks (RNNs) are artificial neural networks that contain at least one loop that allows the network’s internal states to be dynamically influenced by its own previous internal states, in addition to any new external input. Formally, containing loops makes RNNs cyclic graphs; in th
Reinforcement Learning
Reinforcement learning (RL) refers to a process in which an agent (biological or artificial) learns how to behave in its environment by using a simple type of information: reinforcers, which index how good or bad something is. The term RL is used by multiple scientific communities to cover different
The Free Energy Principle
The free energy principle is a mathematical principle that describes how interacting objects or “things” (defined in a specific way) change or evolve over time. In this context, a thing is a set of states that can be meaningfully distinguished from other such things (e.g., particle, person, or popul
The Turing Test
“The Turing test” is the name given to a test of human-level intelligence in machines, invented by Alan Turing, the renowned mathematician, codebreaker, and computer pioneer. In Turing’s imitation game, a human interrogator has text conversations with both a human being and a computer that is preten
Transformers
Transformers are a type of artificial neural network architecture designed to process sequential data—such as text—by selectively focusing on relationships between different elements in the sequence through a mechanism called attention. Unlike earlier approaches that processed sequences one element