YouTube: session was not recorded
live AI Salon / AI 101 session on AI topics.
Next: AI 101 Live Session, 2024-02-01
Previous: AI 101 Open Session, 2024-01-26b
Introduction of Hackam, a note-taking tool; discussion on basic AI tools and tips; advice on using speech-to-text programs; experience with transcription tools; discussion on prompt engineering; organization of GPTs; using Chat Gpt for productivity, creativity, and data analysis; using Chat for generating prompts for image creation; using rephrased prompts with Dolly; using Chat Gpt for text manipulation and experimentation; regular practice and experimentation with Chat GPT; discussion on AI tool that generates traditional Indonesian Batik designs.
Introducing Hackam and Discussing AI Tools
Pete Kaminski introduced a note-taking tool called Hackam, intending to share the tool link with participants. He also discussed basic AI tools and tips, such as improving chat and crafting better prompts. He briefly mentioned the idea of building a GPT. Julia Jack sought advice on a speech-to-text program to use with Gpt on her new Windows computer. Pete Kaminski and R J recommended using the built-in dictation tool on Windows, while R J introduced Julia Jack to 'Speech Notes'. Pete also shared his experience with 'Mac Whisper'.
Deaf Woman Seeks Hearing Aid Advice
Jeannie Bell, who is deaf, sought advice from Pete Kaminski about a headset that could help her hear during meetings. Pete suggested a bone conduction headset made by a company called Shox. He shared his positive experience with the product, noting that it vibrates on the temple to transmit sound. Pete also mentioned that Shox often has sales, offering discounts of up to 30%.
Transcription Tools Discussion
Pete Kaminski discussed his experiences with transcription tools, highlighting Mac Whisper and the script as particularly effective. He expressed his admiration for Mac Whisper's transcription quality and its ability to clean up disfluencies, and he mentioned the script for its ability to edit transcripts to edit the video. He also mentioned the possibility of using the tool in real-time. Julia Jack raised a question about commenting on the meeting prior to its start, which led to a discussion about the functionality of the Mighty Networks app. Jeannie Bell also had a brief interjection.
Prompt Engineering and AI Tools for Productivity and Creativity
Pete Kaminski discussed the significance of prompt engineering, emphasizing its potential to generate substantial income without requiring extensive computer knowledge. He highlighted the role of image generators in prompt construction and the importance of careful prompt placement, especially in stable diffusion. Pete also discussed the benefits of using a tool called Chat Gpt for productivity, creativity, and data analysis. R.J. praised Pete's list, emphasizing the benefits of using AI to organize thoughts and improve sleep. Jeannie Bell shared her experience of using different GPT assistants for different topics to avoid asking 'dumb' questions and to have more in-depth conversations. Pete demonstrated how a system could recognize when a topic changed, suggesting that users should start a new conversation. He also touched upon how GPT can preset the context for a conversation and explained the idea of explaining complex topics to people of different educational backgrounds.
GPT Organization and Note-Taking Tools
Pete Kaminski discussed the organization of GPTs using various methods, including naming them based on their purpose and using tools like Obsidian and HackMD for note-taking. He also introduced a Chrome plugin called "Create Link" and shared his personal use of a chatbot, ChatGPT, to help with motivation and organization. Pete suggested the potential of using ChatGPT to synthesize information from various conversations. Towards the end of the meeting, Julia Jack asked a question which was not included in the transcript.
Chat Image Prompt Generation Discussion
There was a discussion about using Chat for generating prompts for image creation, with a focus on Mid Journey. Pete Kaminski explained his approach, which involves setting up numerous prompts and then queuing them up. He also mentioned that he uses the info command to monitor his current status. He shared some of the prompts he has used, such as "create some problems for something super bright and shiny" and "don't create an image", and showed examples of the resulting images. Julia Jack had further questions about the process, and Pete explained how he uses GPT to help with generating prompts.
Effectiveness of Rephrased Prompts in Image Generation
Julia Jack and Pete discussed the effectiveness of using rephrased prompts with Dolly, a program that generates images based on user input. Pete shared his experience, noting that Dolly often expands prompts and generates what it believes is a better prompt, and these prompts are more deterministic. They agreed on the need for experimentation to determine the effectiveness of using rephrased prompts. Additionally, Pete discussed the use of Chat Gpt for text manipulation and experimentation. He emphasized the importance of practice and experimentation to improve prompting skills, and demonstrated how to edit the prompt to maintain consistency and creativity in the output.
Chat GPT Practice and AI Tool Development
Pete Kaminski emphasized the importance of regular practice and experimentation with Chat GPT for improving proficiency. He suggested using it for various tasks, including language learning, and shared his personal experiences with writing poems in multiple languages. Pete also discussed a network error issue and advised attendees to copy and paste any valuable information. He also shared his work on an AI tool that generates traditional Indonesian Batik designs and has the potential to create lists of traditional fabric and painting designs. The discussion ended with Pete saving the chat and hinting at future meetings to delve deeper into the topic. R J expressed interest in developing local models for conversational tasks to avoid dependence on external platforms, which Pete suggested exploring the Llama file.