The author also contrasts their experience with the AI chatbot to their positive experience with GitHub Copilot, a supervised AI tool. They argue that supervised AI services, where the output is overseen by a trained human, are more reliable and likely to stick around. The author remains hopeful about the future of unsupervised AI services, but acknowledges that they currently still require significant improvements.
Key takeaways:
- The author's team tested an AI chatbot for their documentation but decided to discard it due to its inability to provide accurate and nuanced responses.
- Improving the chatbot's performance would require more/better data and helping users navigate the mistakes that the chatbot made, which was deemed too time-consuming compared to improving their guides.
- Supervised AI tools, like GitHub Copilot, are more successful because they are overseen by a trained human who can spot and correct mistakes.
- Unsupervised AI tools have the potential to revolutionize various industries if they can improve their accuracy and reduce the need for human supervision.