🎯 6 strategies to get better results from chat-based AIs

Juan Delgadillo

Jul 24, 2024 • 3 min read

Welcome back, folks! This week's issue comes with 6 ways to get better, more relevant and accurate answers from LLMs. We hope you enjoy it!

💬 In this week's letter:

OpenAI: 6 strategies for getting better results from Large Language Models (LLMs).

Was this email forwarded to you? Subscribe here!

6 strategies to get better results from LLMs

When asking chat-based AIs—also known as LLMs—for help on your creative tasks, you may not get what you expect. They often make stuff up, have limited reasoning, and struggle to understand complex multistep problems.

So, what is the best way to ask LLMs for help and get answers with higher accuracy and relevance?

OpenAI, the company behind ChatGPT, shared 6 strategies for getting better results from LLMs:

1. Write clear instructions

These LLMs can’t read your mind. If answers are too long, ask for brief replies. If answers are too simple, ask for expert-level writing. If you dislike the format, demonstrate the format you’d like to see. The less the LLM has to guess at what you want, the more likely you’ll get it.

2. Provide reference text

Language models can confidently invent fake answers, especially when asked about esoteric topics or for citations and URLs. In the same way that a sheet of notes can help a student do better on a test, providing reference text to these models can help in answering with fewer fabrications.

3. Split complex tasks into simpler subtasks

Just as it is good practice to break a complex problem into a set of smaller problems, the same is true of tasks submitted to a language model. Complex tasks tend to have higher error rates than simpler tasks. Furthermore, complex tasks can often be re-defined as a workflow of simpler tasks in which the results of earlier tasks are used to construct the inputs to later tasks.

4. Give the model time to "think"

If asked to multiply 17 by 28, you might not know it instantly, but can still work it out with time. Similarly, models make more reasoning errors when trying to answer right away, rather than taking time to work out an answer. Asking for a "chain of thought" before an answer can help the model reason its way toward correct answers more reliably.

5. Use external tools

Compensate for the weaknesses of the model by feeding it the results of other tools. For example, one tool can tell the model about relevant documents. A code execution engine like OpenAI's Code Interpreter can help the model do math and run code. If a task can be done more reliably or efficiently by a tool rather than by a language model, offload it to get the best of both.

6. Test changes systematically

Improving performance is easier if you can measure it. In some cases a modification to a prompt will achieve better performance on a few isolated examples but lead to worse overall performance on a more representative set of examples. Therefore to be sure that a change is net positive to performance it may be necessary to define a comprehensive test suite (also known an as an "eval").

The methods described can sometimes be used in combination for greater effect. We encourage experimentation to find the methods that work best for you.–OpenAI

Want to work together?

Work 1:1 with me – book a coaching or strategy session.
Advertise – showcase your brand, product, or service to our engaged audience of entrepreneurs, creators and investors developing a competitive advantage on AI.

Forward to a friend, sharing is caring.

Anything else? Hit reply to send us feedback or say hello. We don't bite!