The author also noted the cost and time involved in using the AI, stating that it took 26 hours and cost $110 to run the AI on 940 files. He concluded the article by advising readers to thoroughly test and evaluate AI on multiple inputs before using it for large tasks, and to parallelize the work to save time and costs.
Key takeaways:
- Generative AI, specifically GPT-4, can be useful for tasks such as fixing typos and bad syntax in code snippets, and even inferring image captions from context. However, it can also make mistakes such as turning images into links, removing content from long articles, and following instructions even when they don't apply.
- Running AI on a large number of files can be slow and expensive. In this case, running GPT-4 on 940 files cost about $110 and took 26 hours.
- It's important to thoroughly test and evaluate AI on multiple inputs before running it on a full set of data. Also, parallelizing the work can help speed up the process.
- Despite its shortcomings, AI can still be a useful tool when used correctly and with the right expectations.