The author predicts that prompt engineering will soon replace what was formerly known as "deep learning". They believe that while supervised learning, which requires substantial effort to label large datasets, was crucial in demonstrating the efficacy of deep learning, the continuous improvements in making it easier and more effective means there's no reverting to the old ways.
Key takeaways:
- LLMs have popularized zero-shot learning, also known as 'prompt engineering', which is easier to use and more effective than labeling data.
- 'Prompt engineering' can be retrofitted onto traditional machine learning techniques such as text classifiers.
- The author has created a library to retrofit 'prompt engineering' onto machine learning, which can be found at https://github.com/lamini-ai/llm-classifier.
- The author believes that 'prompt engineering' will soon replace what was previously known as 'deep learning', as it is easier and more effective.