AI 101 Live Session, 2024-02-03

YouTube: https://youtu.be/Q300wgEp3GE

Recorded live AI Salon / AI 101 session on AI topics.

Next: AI 101 Live Session, 2024-02-08

Previous: AI 101 Live Session, 2024-02-01

AI Summary (hopefully useful, may be inaccurate)

Quick Recap

Pete Kaminski discussed the use of a collaborative note-taking tool called HackMD. He also touched upon markdown and shared links from previous sessions. Pete then listed several topics he intended to cover, including mid-journey variations, templates, local LLMs, image generators, version control with Git and Github, and GPTs. He also discussed the use of local LLMs and image generators on individual computers, emphasized the significance of Github for professional development, underscored the functionality and benefits of the Llama file framework, discussed his experiences with a software tool, discussed the capabilities of image generators, discussed the process of using Chat Gpd and other image generation tools, demonstrated the use of these tools, discussed the process of using Chat Gpd and other image generation tools, discussed the installation process of certain software, discussed the Focus software, and finally discussed how to get started with AI 101.

Summary

Collaborative Note-Taking Tool Discussion

Pete Kaminski discussed the use of a collaborative note-taking tool called HackMD. He shared a link in the chat for participants to use and explained the three different modes of the tool. Pete also mentioned the use of markdown and shared links from previous sessions. There was a suggestion from R J to replace the term "ephemeral" with "temporary link" to make it more understandable for everyone. Pete then listed several topics he intended to cover, including mid-journey variations, templates, local LLMs, image generators, version control with Git and Github, and GPTs. R J requested that the topic of local LLMs be discussed first due to a prior commitment. Pete agreed to this and suggested they continue to explore the topic further.

Local AI Tools and LLM Updates Discussed

Pete Kaminski discussed the use of local LLMs and image generators on individual computers, emphasizing that these tools require setup, maintenance, and significant computing power. He compared the experience of using local AI with that of being a passenger on an airline, suggesting that local AI can be more complex and intense. Pete also mentioned that certain image generators, like Llama File and Focus, can help manage some of this complexity. He also shared his plan to update a lava lama llama file installed on his system using an old post from ASalon and planned to share a link to expert instructions on setting up LLMs. Finally, he highlighted a blog post from Mozilla about LLMs and noted that the main distribution hub for Llma file is on this site.

Github's Role in Professional Development

Pete Kaminski underscored the significance of Github for professional development, particularly within the AI sector. He highlighted its utility in showcasing contributions and its role in the ecosystem of easy-to-use website-making and information management tools. Pete also shared his personal experience of using Github for maintaining websites and suggested others to consider using it. However, R J brought up his negative experience with Github, being banned for link sharing. Pete clarified that while an account is not necessary for viewing public content on Github, it is required for making changes or adding comments. He also discussed the challenges software developers face when using GitHub, mentioning its technical nature and lack of user-friendly features.

Llama File Framework: User-Friendly Models

Pete Kaminski discussed the functionality and benefits of the Llama file framework. He emphasized its user-friendly nature, allowing users to download and run different models without the need for complex setup or computer-dependent tweaks. He also explained the concept of "models" as different brains or functions, such as language models, image models, and so on. Furthermore, he touched upon the idea of "multimodal" models, which can handle multiple modes, such as text and images.

Software Tool User Experience and Local Server Setup

Pete Kaminski discussed his experiences with a software tool, which he found to be more user-friendly than previous versions. He detailed the process of downloading and running a file, emphasizing that the steps are more intuitive for Mac or Linux users, but may require additional steps for Windows users. Pete also explained how to interact with a local server running on a computer using a specific domain name. He highlighted the utility of this process in case of societal collapse, as it could run independently with minimal requirements.

Chatbot's Image Recognition Capabilities Discussed

Pete Kaminski discussed a chatbot he is using on his computer. He highlighted the capabilities of the bot, which includes image recognition, and demonstrated its ability to analyze an image of a $100 bill, interpreting it as a $20 bill. Pete also compared the image recognition capabilities of different versions of the bot and mentioned the potential for the bot to generate prompts for image descriptions. He also mentioned the bot's ability to resize images and its tendency to generate prompts for Dolly, another image generation tool.

Image Generator Capabilities and Competitors Discussed

Pete Kaminski discussed the capabilities of image generators, mentioning competitors to Llama, text generators, and local image generators. He demonstrated the ability of these systems to enhance prompts and make changes to images without redrawing the entire prompt. He suggested that Chat Gbd is used to expand prompts when Dolly, another image generator, struggles with short prompts. Towards the end, he demonstrated how a short prompt can be used to generate a long one using Chat Gbt.

ChatGPT and Image Generation Tools Demonstration

Pete Kaminski discussed the process of using Chat Gpd and other image generation tools, specifically focusing on the tool called "Draw Things" which uses Stable Diffusion. He explained how these tools take a complex concept or description and condense it into a single image. Pete also mentioned that these tools are open-source and can be run on different platforms like Mac, iPad, and iPhone. He demonstrated the use of these tools by creating images and discussed the settings and controls involved in the process. Towards the end, he humorously commented on the sometimes unappealing aesthetics of the images generated by these tools.

Process Improvement and Tool Experimentation

Pete Kaminski discussed his efforts to improve the speed of a process, which involved reducing the image size and adjusting the batch size. He mentioned that he had just learned about a new tool called 'focus' and planned to experiment with it. Pete also compared two different tools, one for Mac users and one for PC users, both of which are used for image generation. He concluded by noting that all these tools run within a web browser.

Software Installation and Comparison Discussion

Pete Kaminski discussed the installation process of certain software, emphasizing the importance of using an older version of Python due to compatibility issues with new versions. He then shifted the discussion to the Focus software, explaining its features and benefits, including the ability to automatically download models. Pete also cautioned about potential fake websites and the need for careful usage. He further compared Focus with Mid Journey and Chat Gpt, highlighting the unique capabilities of each software. Towards the end, he discussed the idea of converting the topics in the table of contents into check boxes for better organization.

AI 101: Starting With Basics and Future Plans

Pete Kaminski discussed with Scott Arndt about how to get started with AI 101. Pete suggested beginning with the basics and noted that the website he created using Massive Wiki builder could be a useful resource. He also mentioned his plans to have guest teachers share their knowledge in future sessions. Towards the end, Pete thanked everyone for their participation and expressed his excitement for the upcoming topics.