LibGuides: Literature Reviews: Artificial Intelligence (AI) Tools

AI Tools Licensed by UT-Austin

log in with your UT Microsoft 365 account (@austin.utexas.edu) and then your EID to access the UT-Austin licensed version of CoPilot. Your prompts and responses are not retained by Microsoft or used to train AI models, and your information is encrypted.
generates text, code, and images using ChatGPT-4 and DALL-E 3.
not currently approved for use with confidential university data (e.g., FERPA, HIPAA, PCI, IRB).
more information is available on the UT Microsoft 365 AI page.

ChatGPT and Other Large Language Models

The table below summarizes key characteristics of the most widely used LLM AI chatbot tools. This table was last updated in July 2024. A more exhaustive list of generative AI tools that may be of interest to university students and faculty is maintained by Ithaka S+R at https://sr.ithaka.org/our-work/generative-ai-product-tracker/

	ChatGPT	Gemini	Llama
Website	https://chat.openai.com/auth/login	https://gemini.google.com/	https://ai.meta.com/llama/
Access	iOS, Android, Web-application	Cannot run locally. Google Gemini website, through Google app integrations	Can be downloaded and run locally, can be accessed through 3rd party online interfaces
Company	OpenAI	Google	Meta
Cost/Upgrade Version	Basic tier is free and can be accessed without an account. An upgraded version (ChatGPT Plus) is available for $20 per month and requires the user to create an account with OpenAI.	Free but requires Google account	Free
Licensing	Proprietary	Proprietary	Open with some restrictions (https://github.com/facebookresearch/llama/blob/main/LICENSE)
User Privacy	OpenAI Privacy Policy (for non-European users) OpenAI Privacy Policy (for EU) Will collect personal information and provide information to partners, in addition to analysis of user behavior.	Gemini Apps Privacy Hub Uses location, past conversation data to provide responses. Will save and share data with other Google products if linked.	Meta Privacy Policy Requires submission of name and email in order to download the model. Once downloaded, the model can then be run locally without sharing data.
Model Training Set	ChatGPT was trained using the Common Crawl open dataset in addition to resources like Wikipedia, books, and news articles.		Llama was trained with data sources similar to those of other LLMs, but only those with publicly available data that are “compatible with open sourcing”. See https://arxiv.org/pdf/2302.13971.pdf for full details on training data and list of data sources.

AI Scholarly Literature Search Tools

Consensus: this literature search, synthesis and extraction tool draws on citations and open access papers from semantic scholar. It discovers papers, summarizes and extracts data. Consensus also includes a "consensus meter" which will answer a yes/no research question with the percentage of found studies saying yes, no or possibly and integrates with Microsoft CoPilot. A basic free account provides unlimited searches and some CoPilot integration, or you can upgrade to a paid account.
Elicit: this literature search, synthesis and extraction tool draws on citations and open access papers from semantic scholar, with a focus on “empirical research (e.g. randomized controlled trials in social sciences or biomedicine)" to discover papers, as well as summarize and extract data. A basic free account gives you unlimited search, unlimited summaries of 4 papers at once, and other features or you can upgrade to a paid account.
ResearchRabbit: dubbed “Spotify for papers,” this free tool draws on citations and open access papers from semantic scholar to create visualizations of connections between papers and scholars, as well as recommend and summarize papers.