On June 27th, the eleventh session of the "Coaching meets AI" dialogue series took place. We began with an overview of the topics covered so far. The key insights were:
- Use of Bots: Oliver Wünsch demonstrated how bots can be utilized in coaching. There was a discussion on how these bots can be used in dialogue with coaches and coachees.
- Problem with Problem Analysis: A significant point was that problem analysis by ChatGPT often remains too general and lacks depth. It was emphasized that incorporating coaching-specific knowledge is necessary.
- Integration of Coaching-Specific Knowledge: It was suggested to feed ChatGPT with professional coaching knowledge to achieve more specific and useful results.
Session on June 27th:
- Introduction by Annette: Annette presented her insights on pattern recognition by AI. She emphasized that AI works based on algorithms and heuristics to recognize patterns and thus understand language. Traditional rule-based approaches were unsuccessful, which is why machine learning was adopted.
- Training AI: It was explained that AI models are trained through large datasets and powerful hardware. These models work with statistical probabilities to make predictions.
- AI Hallucinations: A central topic was the "hallucination" of AI, where the model generates answers even when it is uncertain. This can be used productively but also critically viewed as it can lead to misinterpretations.
- Custom Instructions: It was demonstrated how to customize ChatGPT through custom instructions and document uploads to implement specific behavior and knowledge.
- Integration of Various Data Types: Annette highlighted that AI can integrate various data types such as text, images, and videos through multimodality to recognize and process complex patterns.
Outlook:
- Research Project: A new research project was introduced, investigating the use of prompt scripts in coaching. Participants are invited to test and evaluate these scripts.
- Next Session: In the next session, the concept of AI coaching with prompt scripts will be further explored and practically tested. Feedback and Requests: Participants responded positively to the session and expressed a desire to delve deeper into the practical application of prompt scripts in future sessions. It was emphasized that custom instructions and AI training should be further explored and refined.