Start main page content

Educational Explainer

DEFINITIONS

What is Fundamental AI Research?

Fundamental AI research involves using ideas from a range of fields such as mathematics, statistics, computer science, and cognitive science to improve AI models and our understanding of how they work. These improvements may involve developing new capabilities, improving efficiency, enhancing safety, or making the outputs more explainable.

What is Artificial Intelligence (AI)?

Coined by Stanford University Professor John McCarthy in 1995, AI refers to building machines that are capable of tasks typically associated with intelligent behaviour, such as problem-solving, decision-making, pattern recognition, natural language processing, and perception.

The core of modern AI is driven by machine learning, where algorithms learn from data to improve their performance over time.

What is Machine Learning (ML)?

ML is a branch of artificial intelligence (AI) focused on developing systems that can learn from data and improve their performance over time without being explicitly programmed.

In ML, algorithms identify patterns within data and use these insights to make predictions, recognize objects, translate languages, or perform other specific tasks. The core idea is that as more data becomes available, ML models become better at recognizing patterns and making accurate predictions.

What are neural networks?

Neural networks are computational models loosely inspired by the structure and function of the human brain and designed to process complex data patterns and recognise relationships in data. They are composed of interconnected "neurons" organised in layers that hierarchically process information, allowing for advanced tasks like image recognition, language processing, and data analysis.

Deep learning refers to neural networks organised with multiple layers of these neurons.

What are Large Language Models (LLM)?

LLMs are large artificial neural networks which are focused on understanding and generating human language. They are trained on vast datasets to be able to predict the most likely next word and can perform tasks like text generation, translation, and summarisation.

ChatGPT and similar models are examples of LLMs that can generate human-like text responses to input instructions, or prompts.

What is Generative AI (GenAI)?

GenAI refers to ML models that can create new content, such as text, images, audio, video, or code, by learning enough about the structure of the data used to train it to produce new examples.

These models, like ChatGPT (OpenAI’s Generative Pretrained Transformer chatbot), use deep learning techniques to generate content that mimics human-like creativity. ChatGPT has been trained on hundreds of gigabytes of text scraped from the internet, which gives it a good enough understanding of human text to respond to prompts it has never seen before.

How does our understanding of human and animal intelligence inform machine intelligence in building AI?

Professor Benjamin Rosman, Director of the MIND Institute says many ideas in AI have drawn inspiration from understanding phenomena in natural intelligence. “For example, artificial neural networks are loosely inspired by abstractions of the way that neurons operate in brains,” he says.

Some other examples are:

  • Convolutional Neural Networks (CNNs) are kind of neural network inspired by the visual cortex, part of the brain that processes visual information. CNNs are specialised for processing grid-like data such as images and capturing spatial hierarchies.
  • Reinforcement Learning is a machine learning paradigm for learning to make decisions. Based on behavioural psychology it models learning through trial and error, using rewards and punishments to shape behaviour.
  • Genetic Algorithms are a class of optimisation and learning procedures, inspired by natural selection. These algorithms evolve solutions over generations using processes like mutation and crossover.
  • Swarm Intelligence is another class of optimisation procedure. Modelled after the collective behaviour of social insects, it solves problems through decentralised, self-organised systems.
Share