OC logo
Libraries Home Start Researching Subject Guides Services About Us
Skip to Main Content

Generative AI-Tools

What is generative AI?

Traditional AI systems are trained on large amounts of data to identify patterns, and they’re capable of performing specific tasks that can help people and organizations. But generative AI goes one step further by using complex systems and models to generate new, or novel, outputs in the form of an image, text, or audio based on natural language prompts. 

Large language Models (ChatGPT, Bing AI, Bard, BLOOM, etc.) have been trained on massive quantities, terabytes, of text and code repositories. By learning associations between words in that training data, they can generate answers to natural language prompts and questions.

Adapted from GitHub

Crafting Effective Prompts

[Context] + [specific information] + [intent] + [response format] = perfect prompt

 

1 - Be Specific: If you ask a vague question, you are likely to get a vague answer. The more details you provide, the better the model can give you what you're looking for. Instead of "tell me about all dog breeds that exist,” ask "What are the different breeds of small dogs suitable for apartment living?"

2 - State Your Intent: If there's a specific purpose for your question, state it in the prompt. For example, instead of asking “explain quantum physics” you could say "I'm helping my fifth-grade son with his science homework. Could you explain quantum physics in a simple way?"

3 - Use Correct Spelling and Grammar: While the model can often interpret and correct spelling and grammar mistakes, providing clear and correct prompts helps ensure you get the best response.

4 - Direct the Output Format: If you want the answer in a specific format, state it in your question. For example, you could ask "Could you list the steps to bake a chocolate cake?" or "Could you explain the process of baking a chocolate cake in a paragraph?"

5 - Ask Follow-Up Questions: If the response wasn't what you expected or if you need more information, ask follow-up questions to clarify.

6 - Experiment with Different Phrasings: If you're not getting the response you want, try asking the question in a different way. The model's results vary based on your input.

7 - Fact-Check: You can ask the model to fact-check its answer but this is not always effective, so be prepared to verify independently.

Adapted from App of the Day

Generative AI- What's It Good For?

Important Warnings about Generative AI

This technology is still in its early days and there are important issues that users should know. 

  • Generative AI produces misinformation, and it can be hard to detect. As OpenAI describes this limitation, "ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers."  This is known as "hallucinating." Generative AI often cannot detect its own errors.
  • There are privacy, accountability, and transparency concerns, regarding training dataset, user information data, and more. This is a topic that is constantly and rapidly changing, with international regulators, ethicists, lawmakers, and many others creating guidelines, rules, and laws.  
  • There is recurring evidence of both subtle as well as overt bias in the output of generative AI-tools. This is an ethical concern for the creators, but it is also something users need to be aware of. If you encounter this, use the built-in reporting feature.
  • Your input becomes part of the available training data available to the company who owns the AI-tool you use. Be aware that you are providing free labor (and this is not the only way that happens!).  
  • There are pending legal actions regarding the IP (intellectual property) of the data used to train various AI-tools. There is value in consciously applying your ethics to considering  the implication of benefiting from stolen content.

Examples of AI-Tools