Generative Artificial Intelligence Guide

Generative AI tools are continuously being developed. You will come across different types, including:

• Text generation (e.g. Microsoft Copilot)
• Image generation (e.g. Microsoft Image Generator)
• Code generation
• Research tools

Gen AI tools take user input known as "prompts" and outputs text, images, and code. They are useful for summarising a topic and brainstorming. Note: their limitations include hallucinations, as well as being inaccurate, out of date, and biased.

Below are some current examples of widely-used tools.

Microsoft Copilot 

Microsoft’s AI chat assistant is built into Windows and the Microsoft Edge browser. It uses Bing search results to generate text. A Large Language Model (LLM) to answer user prompts, it has limitations of inaccurate, out of date and bias in its results.

https://copilot.microsoft.com/ 

ChatGPT

One of the first chatbots. Using natural language prompts, ChatGPT generates text answers using LLM. However, the data it uses can be out of date and its well-known for imagined responses. 

https://chat.openai.com/

Google Gemini

Gemini is Google's chatbot -previously known as Bard. It uses a LLM to answer queries. Prompts can be text, images or code. It uses current data from online searches in its database and has limitations of inaccurate, out of date and bias in its results.

https://gemini.google.com/

Perplexity

Perplexity can use a variety of media input, but focuses on accuracy over user-friendliness and tends to be less conversational. It uses current information and cites sources.

https://www.perplexity.ai/ 

 Riff

The Riff Reflection tool generates questions to help you think about a subject more deeply, and from new perspectives.

https://riffbot.ai/

Microsoft Image Creator 

Microsoft. (2024). Designer image creator. https://designer.microsoft.com/image-creator.
Prompt: Historical black and white photo of women suffragettes in London 1910.

Microsoft Image Creator Produces high-quality images uses natural language prompts. It has limitations of distortions and bias.

https://designer.microsoft.com/image-creator

Canva AI Image Generator

Canva. (2024). Canva image creator. https://www.canva.com/ai-image-generator/.
Prompt: Mountains at sunset.

Canva uses Magic Media, text to art and video generator for uploading to Canva documents. Social media posts, presentations, infographics, video and many more are some of the available templates. 

https://www.canva.com/ai-image-generator/ 

Dall-E 2 

Dall-E 2 generates editable, high-quality images from natural language input. There are ethical concerns about how it was trained and it cannot process more complex prompts.

https://openai.com/dall-e-2 

Github Copilot

Supports programmers to write code quickly the results of which - as with other Generative AI tools - need to be assessed critically.  It has more developed coding functions than ChatGPT.

https://github.com/features/copilot

Text generation tools such as Microsoft Copilot, ChatGPT, and Google Gemini can also be used for coding. 

ChatPDF  

Very easy to use, drag and drop your PDF and a summary is generated by answering possible questions you may have asked in reading the content.

https://www.chatpdf.com/

Elicit 

Elicit works with Semantic Scholar articles to highlight research in a field of study based on a research question. From top papers, information is summarised into main concepts.

https://elicit.com/

OpenRead AI

OpenRead AI offers a natural language question platform and provides a summary and a list of articles on your topic, OpenRead summarises PDF's that you upload.

https://www.openread.academy/

ResearchRabbit  

Research Rabbit integrates with Zotero, the reference management system, where you can upload a collection of articles from Zotero to Research Rabbit and it will summarise the content.

https://www.researchrabbit.ai/ 

Scite

Scite extracts sentences around an in-text citation, giving the reader important context on the citation as well as coding the reference as 'supporting', 'mentioning' or 'contrasting'. It is distinct from other scholarly Gen Ai offerings and is reliant on partnerships with publishers for access to publications.

https://scite.ai/

SciSpace 

Good for a general review of a topic on an introductory level. The SciSpace function offers dynamic summaries of uploaded PDF's. It can have innacurate summaries or citations.

https://typeset.io/ 

Scopus AI tool

Scopus is an established reputable database for scholarly material that links citation details with authors and institutions. Scopus launched a Gen AI tool that uses natural language questions to create a summary of the research published on that topic, referencing key articles published in sources published in the past five years and generated with links to authors and key subjects.  

Semantic scholar  

Semantic Scholar summarises text with snippets of important information from articles and citation metrics. Semantic scholar holds less scholarly material than google scholar, it is reliant on partnerships with publishers for access to publications.

https://www.semanticscholar.org/

Microsoft. (2024). Designer image creator. https://designer.microsoft.com/image-creator.