Artificial Intelligence Knowledge Hub

Definitions

  • Artificial Intelligence (AI) is a broad term for digital technology that can perform tasks that require "intelligence", such as reasoning, making decisions, learning from mistakes, communicating, and problem solving.
  • An Algorithm is a set of instructions or rules that a computer (including smartphones) uses to complete a task.
  • Generative AI is a type of AI that generates text, images, audio, video, or other media based on user prompts using machine learning. Chatbots are a common example of generative AI.
  • Hallucinations refer to responses from large language models (LLMs) such as ChatGPT that seem plausible but are inaccurate. An example would be a citation for a book that doesn't exist.
  • Large Language Models (LLMs) are a foundation model, which is a type of machine learning model trained on vast amounts of data. LLMs are trained on massive amounts of text to carry out language-related tasks. Google Translate and ChatGPT are both examples of LLMs.
  • Machine Learning is a type of AI that uses algorithms to 'learn' without all of it's instructions being explicitly programmed. Examples include LLMs (see above), virtual assistants like Alexa, and facial recognition.
Further resources:

Guidance and Statements

Data Protection and Ethics
Data Protection Bias and fairness Environmental impact

Examples of AI Tools

  • ChatGPT is a large language model (LLM) chatbot developed by OpenAI. GPT = Generative Pre-Trained Transformer. It is trained on a massive dataset of text and code and can generate human-like text in response to a wide range of prompts and questions. It is not connected to the internet and can make stuff up to fill gaps in its ‘knowledge’.
  • MS Copilot (previously Bing AI) also uses GPT-4 (a LLM) to answer questions in a conversational way. It will reference its sources which include: The internet, it’s own knowledge base and conversation history. Copilot can generate images using Dall-E 3, a text-to-image generator created by OpenAI. Images generated, such as anatomical images, may not necessarily be accurate.
  • Gemini (Previously Google Bard) is a more powerful and versatile LLM than GPT. It will answer questions in a similar way but it is trained on a dataset that includes both text and code (GPT is trained on text only). It also has access to real-time internet while GPT does not.
  • Elicit is a little different and uses machine learning and natural language processing (NLP) to help you find relevant research papers. It will then summarise the paper and extract key information such as formulas or statistical tests. It is a good alternative to using a clinical database (such as medline) if you want to find papers quickly.
  • Consensus also uses machine learning to identify research papers that will answer a specific research question. Based on the result of the papers it will provide a ‘consensus’ answer to your research question. It only searches for published research via Semantic Scholar so it is more reliable that the GPT tools. There is currently a 6 month lag of data due to indexing - this is good as there is indexing.
  • Humata uses NLP to analyse text in a document/research paper. You can upload a PDF and ask it questions such as ‘what are the results of this trial’ or was NNT used? It will also generate new writing (rewords) based on existing documents. 

AI in Health

Keep up-to-date

KnowledgeShare is our targeted current awareness service. Alerts are emailed out every fortnight. Your KnowledgeShare profile can be created to be as broad or specific as you need, and can be updated if your interests change.   It can include non-clinical topics such as AI, leadership and research.

Library & Knowledge Service

Facilitating evidence-based decision making at East Cheshire NHS Trust
Copyright © 2023