Today, there are many AI products that are available for free, and the number of products continues to grow exponentially. This overview provides brief descriptions of the various types of AI products available for use.
The purpose of this overview is not to encourage or discourage use, but rather to inform and educate the MICA community.
AI Products for Research
The following products are generative AI tools developed by non-profit research organizations for optimizing and accelerating particular research methods and activities. These products are trained on a variety of sources, some more academic than others. Below is a table of a small selection of these products, adapted from Georgetown University Library and University of South Florida Libraries:
Product | Description | Training Data | Is it free? |
Elicit | Using large language models (LLMs), Elicit finds academic papers relevant to your topic by searching through papers and citations and extracting and synthesizing key information. Initiated in 2015 by the Allen Institute for AI. | Semantic Scholar Database | Free trial available. Pay for credits after trial expires. |
Semantic Scholar | Semantic Scholar (which supplies underlying data for some of the other tools on this list) provides brief summaries ('TLDR's) of the main objectives and results of papers. | Semantic Scholar Database | Semantic Scholar is currently free. |
Research Rabbit | Research Rabbit is a citation-based mapping tool that focuses on the relationships between research works. It uses visualizations to help researchers find similar papers and other researchers in their field. | Research Rabbit uses multiple databases, but does not name them (more information can be found on the FAQ page). | Research Rabbit is currently free. |
Perplexity | Using LLMs, Perplexity is a search engine that provides AI-generated answers (much like ChatGPT) including citations which are linked above the summaries. | Internal search index | Free with paid subscriptions available. |
Scholarcy | Scholarcy summarizes key points and claims of articles into 'summary cards' that researchers can read, share, and annotate when compiling research on a given topic. | Scholarcy only uses research papers uploaded or linked by the researcher themselves. It works as a way to help you read and summarize your research, but is not a search engine. | Free (short articles only); Paid version allows articles of any length. |
Text Products
The following products are generative AI tools which can only generate text content. Some of these products are already integrated into other platforms and application you may currently be using.
As always, these products are made to produce human-like communication and text, though they are not actually intelligent. Text generated by these products may not necessarily be true despite being trained on potentially true and reliable sources. Additionally, unlike the research products shown above, these products can generate "hallucinations" when asked to provide real information such as a citation, meaning incorrect or nonexistent sources.
Below is a table of a small selection of these products, adapted from Georgetown University Library and University of Illinois Urbana-Champaign Libraries:
Product | Description | Training Data | Is it free? |
Microsoft Copilot | This LLM is a chatbot created by Microsoft and is incorporated into the search engine Bing and can also be incorporated into Microsoft 365 software. Users can interact with Microsoft Copilot through Microsoft’s web browser, Edge, by typing a prompt into Bing’s search engine, and clicking the Chat button at the bottom of the search dropdown menu. | Data from Microsoft apps such as Word, Excel, and Powerpoint and the Microsoft Graph (data from email, calendar, files, chat) | Copilot is free with basic access. Paid versions include Copilot Pro and Copilot for Microsoft 365. |
ChatGPT | Chat Generative Pre-trained Transformer, or ChatGPT, is a chatbox developed by OpenAI that uses large language models to craft text. Launched in November 2022, ChatGPT is free and available after creating an account with OpenAI. | The paid versions of ChatGPT are currently connected to the internet through Bing. The free version was trained on data last updated in September 2021, but that might change in the future. | ChatGPT-3.5 is the free model, but the latest GPT-4o is paid. |
Gemini |
Designed by Google, Gemini (formerly Bard) is an AI-powered chatbot that responds to natural language queries with relevant information. As with ChatGPT, researchers can use Gemini to aid in topic development and initial source discovery. You may already notice Gemini in your Google search results when it provides an "AI Overview" in response to your search. These are usually, though not always, prompted if you type a question into the Google search bar. |
Gemini can currently connect to the Internet. | Gemini is currently free to use. (Google account required, including MICA Google accounts.) |
Image Products
The following products are generative AI tools which can only generate image content. Unlike many of the text generating products above, these image generative AI products have been less transparent about their training data.
Many of these products rely on an automated process called scraping in which digital media is detected and extracted from websites automatically using a programming language like Python. Artists, designers, and other creators who make and share their work on the Internet have spoken out against scraping as an unethical practice that uses an individual's intellectual property and copyright protected work without their consent. For more on the ethical concerns of this practice, see Ethics, Concerns & Limitations.
Below is a table of a small selection of these products, adapted from Georgetown University Library and University of Illinois Urbana-Champaign Libraries:
Product | Description | Training Data | Is it free? |
DALL-E 2 |
DALL-E 2, developed by OpenAI, is a generative AI tool that generates images from text descriptions. It can apply artistic concepts and styles to generative images based on the input. Similar to GPT4, DALLE-3 is an updated and improved version that is available to ChatGPT Plus and Enterprise users. Microsoft has incorporated DALL-E 3 into Bing’s Image Creator tool. |
OpenAI has not shared specific sources of its training data for products like DALL-E. However, there is evidence that many image generative AI tools have scraped large quantities of images from the internet for training their models. |
Users must purchase credits to generate images through prompts, edit requests, or request image variations. The current base rate for credits is 115 credits for every $15. Pricing information is available on OpenAI's Pricing page in the image models section. |
Midjourney | Midjourney is a generative AI program that produces images based on text and other images. Midjourney Inc. designed the tool using text-based prompts with Discord bot commands. |
Midjourney has not shared specific sources of its training data. However, there is evidence that many image generative AI tools have scraped large quantities of images from the internet for training their models. |
All models are subscription-based. Visit Midjourney's Subscription Plans page for information. |
Stable Diffusion |
Stable Diffusion is a text-to-image open-source generative AI tool created by Stability.Ai. Stable Diffusion offers a free membership model for non-commercial and personal use, as well as paid professional and enterprise memberships. |
Stable Diffusion is built on datasets collected by LAION and derived from non-profit Common Crawl, which have scraped billions of webpages for images. | All models are subscription-based. Visit Stable Diffusion's Pricing Plans page for information. |