Photo of a robot writing with a pencil.

AI Apps: Text Generation

With the introduction of ChatGPT in November of 2022, AI chatbots have become part of the conversation in the world of education and technology. Generative AI such as chatbots run on a large language model (LLM) which helps the program craft human-like responses and respond to diverse queries effectively. For example, the LLM powering ChatGPT is a generative pre-trained transformer (GPT), which is rapidly evolving in the breadth of its capabilities with each new version that has launched since late 2022. However, it’s important to recognize that these language models have their limitations, including potential biases, inaccuracies, and reliance on predetermined data sets.

To gain a quick overview of ChatGPT, and three of its competitors: Google Gemini, Microsoft Copilot, and Claude, refer to the table below. For a more in-depth analysis of each tool, continue scrolling down this page. During your exploration, learn more about how to write productive prompts to get the most of these tools.

Subscription Model Distinctions Limitations Citation Capability
ChatGPT Free (GPT 3.5 & Limited 4o)
  • Temporary chat option (your data won’t be saved and used for training)
  • Can upload files and images (available to free users on limited basis through 4o)

Plus only features:

  • Manage memory of your data (gradual rollout to plus accounts beginning Feb. 2024)
  • Image generation
  • Custom GPTs
  • Timestamp January 2022
  • Prone to being “at capacity” which prevents access
  • 4o access is available on limited basis
  • No access to custom GPTs
  • Struggles to reliably cite sources. Can hallucinate and/or provide broken links to cited resources.
Plus (GPT 4 & 4o)
  • Timestamp April 2023, but can leverage search engine
Microsoft Copilot Free

(access with Montclair.edu credentials)

  • Upload images
  • Image Generation
  • Chat with voice
  • Toggle between output tone styles
  • Utilizes GPT 4
  • No chat history saved (when using Montclair account)
  • Cannot upload files (only images can be uploaded)
  • Enhanced transparency when citing sources. Fairly consistent when providing links to cited sources.
Claude Free
  • Upload files (including images)
  • More safety guardrails in place to prevent harmful behavior (i.e., Constitutional AI model)
  • Timestamp August 2023; cannot access current information from the web
  • Cannot generate images
  • Prone to being “at capacity” which prevents access
  • Enhanced transparency when citing sources. Very consistent when providing links to cited sources.
Google Gemini Free

(access with personal Google account)

  • Upload images
  • Chat with voice
  • Leverages Google Search
  • Quick edits to output (i.e., “modify this response,” and toggle between drafts of output)
  • Conversational in nature, so may struggle to execute more formal writing tasks.
  • Cannot upload files (only images can be uploaded)
  • Can “double-check” responses with Google Search
  • Can provide citations once prompted

ChatGPT

ChatGPT is an artificial intelligence (AI) tool able to produce written work that sounds convincingly human-produced. The AI technology powering ChatGPT is a generative pre-trained transformer (GPT), which is rapidly evolving in the breadth of its capabilities with each new version that has launched since late 2022. These versions are referenced numerically, with GPT-3, GPT-3.5, GPT-4, and GPT-4o being the most recent.

OpenAI offers free and subscription options for ChatGPT users. The free version utilizes GPT-3.5 and offers limited access to GPT-4o the subscription platform, ChatGPT Plus, utilizes OpenAI’s newest language model, GPT-4 and 4o, providing users choice between models. Some strengths ofGPT-4 are: greater accuracy compared to GPT-3.5, broader general knowledge, improved safety behavior, capability of analyzing longer contexts, ability to accept image and file input, and ability to generate images utilizing DALL-E (OpenAI, 2023). GPT4o, introduced in May of 2024, is similar to GPT-4, however, it processes information quicker and has multimodal integration, making it well suited for real-time language translation and speech recognition (Henry, 2024).

Custom GPT features are available for Plus subscribers which enable the ability to utilize a library of third-party tools such as Expedia, Wolfram, Zapier, and Canva within a chat. Plus subscribers can also create their own customized GPTs to enhance the user experience and optimize accuracy of ChatGPT for specific tasks. GPT-4 and 4o still utilize a dataset that is limited to events prior to April 2023 but can leverage its search engine (Bing) to bypass this limitation.

Microsoft Copilot

Microsoft Copilot is both a stand-alone text generator in addition to an Office 365 workspace integration and utilizes the ChatGPT 4 large language model.

Like Google Gemini and ChatGPT Plus, Copilot is able to integrate search results from its engine to supplement the timestamped shortcomings of its dataset. Additionally, it offers the functionality to upload and generate images, chat with voice, and users can toggle between output tone styles.

You can access the browser-based version of Microsoft Copilot with your Montclair credentials; as of April 2024, this version of Copilot does not offer functionality to upload files other than images when crafting a prompt. However, the mobile Copilot application does have this functionality and users can download and access the app without logging in with a Microsoft account from a mobile device.

Claude

Claude made headlines in March of 2024 when introducing the latest version of its large language model, Claude 3, which outperformed GPT-4 and Gemini Ultra on a number of benchmark tests (Anthropic, 2024). Claude, built by Anthropic, prioritizes safety, and has created a Constitutional AI model to help fine-tune its training. While Claude does not have access to a search engine and functions with a timestamp of August 2023, users do have the ability to upload files to add context to their prompts. Additionally, Claude is fairly accurate when generating citations and is transparent about the sources it utilizes in its output without needing to be explicitly prompted to do so. Claude is a free tool; create an account utilizing your institution or personal credentials.

Google Gemini

Google Gemini is Google’s large language model and AI text generator. Similar to Microsoft Copilot, Google Gemini is both a stand-alone text generator as well as an integrated Google Workspace tool. Google Gemini is able to leverage its search engine to reference current information in its output and can generate images based on a prompt. Gemini offers the ability to make quick edits to output that include modifying the tone and length of responses generated along with the ability to toggle between multiple drafts of output. Overall, it is more conversational in nature than ChatGPT however, as of April 2024, users are currently unable to upload files (other than images) to provide prompt context.

Gemini is currently unavailable for activation on Montclair Google accounts. You can access it by authenticating with your personal Google account in Gemini for your browser, or through the Google mobile app.

 

References

Anthropic. (2024, March 4). Introducing the Next Generation of Claude. Anthropic: Announcements.

OpenAI. (2023). GPT-4 is OpenAI’s most advanced system, producing safer and more useful responses. OpenAI: Research.

Henry, W. (2024). GPT-4 vs GPT4o: A Comprehensive Comparison and Analysis. Medium.