What I learned from creating an AI of myself
๐ Abstract
The article discusses the author's experience of creating an AI assistant, "SarahGPT," to efficiently and compassionately address diversity, equity, and inclusion (DEI) challenges in the workplace. The author's goal was to make the knowledge and expertise from her book on DEI more accessible and practical for users.
๐ Q&A
[01] Creating SarahGPT
1. What was the author's primary goal in writing the book "Diversity, Equity and Inclusion: How to Succeed at an Impossible Job"?
- The author's primary goal was to get the book in the hands of as many people as possible and provide them with a practical, easy-to-access resource on DEI.
2. What were the limitations of a print book that the author wanted to overcome?
- The limitations of a print book include:
- Users need to have the book with them or remember the content
- Users need to be able to connect the challenge they are trying to solve with the relevant chapter
- Users need to have the time and patience to reference the book on the go
3. How did the author create SarahGPT to address these limitations?
- The author used the content from her book and other writings (totaling 144,661 words) to train an AI model, similar to ChatGPT, to create SarahGPT.
- The author wanted to ensure SarahGPT had a tone and style that was more authentic and empathetic, replicating the conversational tone used in her book.
- The author also designed a unique brand and visual identity for SarahGPT to reflect her own persona and the purpose of the AI assistant.
4. What were the key steps the author took to create and refine SarahGPT?
- The author uploaded her written content to train the AI model
- She worked to ensure the tone and style of SarahGPT matched her desired conversational and empathetic approach
- She created a unique brand and visual identity for SarahGPT
- She tested SarahGPT extensively, refining the training and prompts to ensure accurate and helpful responses
5. What are the author's thoughts on the potential of custom-built AI models like SarahGPT?
- The author is optimistic about the ability for anyone to create their own GPT models with a specific purpose and use case, as it unlocks opportunities to educate them with highly specified information and achieve a more authentic tone.
- The author believes that as we develop and utilize custom-built AIs in our daily work, instead of one-size-fits-all models, AI will make our lives much easier and better.