Google

An example illustrating the limits of prompting involves asking a Large Language Model to translate words from Inuktitut to English. If the model's pre-training did not include Inuktitut data, it will have a poor understanding of the language, and prompting alone will not enable it to perform the translation task effectively.

Example of Prompting Failure: Inuktitut Translation

Based on the provided scenario, identify the fundamental reason for the chatbot's poor performance and explain why the proposed solution is unlikely to succeed.

Specialized Chatbot Failure

A software developer attempts to use a large language model to generate code in a rare, esoteric programming language from the 1970s. The model was pre-trained on a vast dataset of modern, widely-used programming languages but not this specific one. Despite trying several detailed and well-structured prompts, the model consistently fails to produce valid code in the requested language. Based on the principles of how these models function, analyze the fundamental reason for this failure.

Obscure Programming Language Task

An advanced AI system is designed to identify plant species from photographs. It was developed and trained using a comprehensive dataset of plants found exclusively in Europe. A botanist working in the Amazon rainforest submits a clear, high-quality photograph of a newly discovered local orchid species. The system incorrectly identifies it as a common European flower that shares a similar color. The botanist wants to improve the system so it can correctly identify this new orchid in the future. 

Learn Before

Related