Which generative AI tool for your task?
People often ask, which AI tool can I use to do X? This aims to help.
Version 1, November 2024
nicolehennig.com

by Nicole Hennig

Tasks in this document
In this document, I offer suggestions about which generative AI tools to begin with for each task.
I don't aim to be comprehensive (there are far too many tools available), but just to suggest some starting points.
1. Search and summarize web results
Language model + search results = more current information, less hallucination
Things to know
Both Perplexity and Copilot search the web and summarize the results.
They link to the sources so you can easily find the most relevant websites for your search.
  • As of Oct. 31, 2024 ChatGPT paid versions now have a web search feature (similar to Perplexity).
  • ChatGPT free has limited web searches per day.
  • Claude cannot search the web.
2. Summarize & ask questions of large documents
Work with documents you upload.
Start with these:
  • Claude (free or paid)
    200,000 tokens
Others:
Context windows compared
Context window = total input and output of a single conversation.
Things to know
You can upload and work with documents on ChatGPT, but the context window is smaller, so you may need to break your document into smaller chunks if it is very large.
3. Work with up to 50 of your own documents at once
Create study guides, outlines, summaries, an FAQ, and "Audio Overview" from a set of your own documents.
Things to know:
  • NotebookLM's Audio Overview will generate a "podcast" conversation between two voices, based on your documents.
  • In Perplexity Spaces, you can search across both uploaded files and web content simultaneously. You can give it specific AI instructions and personas. You can opt out of training in your settings. See A Student's Guide to Using Perplexity Spaces.
4. Analyze your data
Upload a .csv file, summarize the contents, ask questions of it, generate tables and graphs, convert it to HTML/CSS & more.
Start with these:
(limited in free accounts, use $ accounts if possible)
Others:
Things to know
  • Free ChatGPT has very limited access to data analysis. You'll run out quickly and have to come back the next day.
5. Get help with coding
Generate code, test the code, evaluate, change and improve code. Download the result.
Start with these:
  • Claude (use the Artifacts feature)
Others:
Things to know:
  • ChatGPT Plus ($) and Claude are both considered to be very capable with coding.
6. Share a link to your conversation
Send a language model's output to others. Handy for demonstrations or showing your work.
These can share a conversation:
(Look for a share button)
No share available:
  • Claude (but you can copy a response and paste it elsewhere)
Things to know:
  • In ChatGPT you can share anonymously or with your name included, see Shared links FAQ from OpenAI.
7. Create custom chatbots
Create a chatbot that converses in a style of your choosing and uses your own prompt for a particular purpose.
Start with this:
Free chatbot builders
(You can define the personality only, no document upload).
Paid versions
Things to know:
  • Free ChatGPT accounts can use custom GPTs, but not create them. (same with Gemini)
  • GPTs allow you to upload specific documents for the GPT to use to answer questions.
  • All of these chatbots (or GPTs) require the user to sign in to at least a free account to use them.
  • There are many other tools for creating chatbots based on genAI, and with those you can embed them into your website and not require users to log in to use them.
8. Create public "pages"
A feature of Perplexity that allows you to curate a set of answers with links to sources.
Things to know:
  • You can turn your research into publicly shareable content, with links to sources.
  • You can control the layout, add sections, add images.
  • You can select your target audience.
9. Describe images with computer vision
Upload an image:
the model can describe it, suggest its context, suggest visual improvements & more.
Start with these:
Others:
Generating alt text for images
A useful prompt:
Please look up best practices for alt text and offer three alternatives for this image.
Things to know
  • Gemini can't describe images of people.
    "Sorry, I can't help with images of people yet."
10. Talk with a voice assistant in the mobile app
Talk out loud to the model and it talks back to you.
Things to know about Advanced Voice Mode in ChatGPT:
  • Choose from various voices in the settings.
  • You can interrupt it.
11. Generate images
Generate images in different styles based on text that you input, or an image that you upload.
Start with these:
  • DALL•E 3 in Copilot or Microsoft Designer
Others:
Things to know
  • Free ChatGPT is limited to 2 image generations per day.
  • Ideogram is pretty good at generating text within an image (like a sign).
  • Midjourney is considered the most capable model, but it's not free. Plans start at $10/month). You can try up to 25 images in a free trial.
  • Claude does not offer image generation.
12. Generate videos
Start with these for text-to-video or image-to-video:
Others in this category:
Start with these for video avatars:
Things to know
  • Most of these aren't free, but many have free trials.
  • OpenAI has announced their video generator, Sora, but it's not live beyond testing with a few creators.
  • Meta has announced MovieGen, but it's not live yet.
  • HeyGen can translate a video of a person talking to another language with lip-syncing to match the new language.
13. Generate voices
Generate spoken audio in different styles based on text that you input.
Start with this:
  • Eleven Labs
    Free version offers about 10 min of audio per month. Paid versions offer more and begin at $5/month.
Others:
Things to know
  • ElevenLabs voice generator can deliver human-like speech in 32 languages.
  • It offers text-to-speech, speech-to-speech, text to sound effects.
14. Generate music
Generate music in different styles based on text that you input.
Start with these:
Things to know:
  • If you are a paying subscriber to Suno, then you own the songs you generate. If you are using a free version of Suno, they retain ownership of the songs you generate, but you're allowed to use those songs for non-commercial purposes.
15. Understand privacy
ChatGPT and other similar tools do not directly learn from and memorize everything that you say to them.
Once trained, the model remains static and unchanged—sometimes for months or even years.
So how do they use your input if you don't opt out?
Targeted Improvements: Companies may focus on areas where the model shows weaknesses or receives frequent user corrections.
Safety and ethical training: User interactions help refine the model's ability to handle sensitive topics and adhere to ethical guidelines.
These kinds of learning from your data can help train future versions of a model.
16. Privacy settings
Individual accounts on these models don't train future models on your input by default:
Other models:
  • Meta AI
    They train on public posts you've made in Facebook, Instagram, and Whatsapp. They don't train on your private posts or conversations. Opt out here and learn more here.
Things to know
Your input isn't used to train future models in these situations:
  • Enterprise, edu, & team versions of ChatGPT
  • Apps built on the API from OpenAI
Temporary chats in ChatGPT
You can start a Temporary Chat by tapping the version of ChatGPT you’re using at the top of the chatgpt.com and selecting Temporary chat.

These won’t be used to improve their models.
Created by Nicole Hennig
  • Date created: October 24, 2024
  • Last updated: November 22, 2024
Feel free to contact me with suggestions or corrections at [email protected].
My website: nicolehennig.com