01:19:12 Grant T.: I have a question for when it’s time; related to VMware vs not VMware. 01:21:11 Sean T.: I have some suggestions for how to use local LLMs to benefit SLUUG 01:23:49 Peter: I'm not hearing Alex's sound. Is everyone else hearing him? 01:23:56 Sean T.: yes 01:24:00 Jamesh: Yes, I hear him 01:24:02 tom chapin: yes 01:24:04 Peter: ok thanks 01:24:08 Peter: I'll check my settings 01:31:40 UN: rag gen ai 01:36:53 Sean T.: Sora is crazy if real 01:37:52 Collard Family: yes 01:37:55 Jamesh: Yes have seen it 01:38:03 Patrick Gronemeyer: ya 01:38:05 Jamesh: beautiful with strange mistakes :) 01:39:54 Jamesh: The one with the cat and the owner has an interesting mistake. 01:40:40 Jamesh: coyote puppies generating out of the ground 01:41:52 Jamesh: Doesn't need a physics engine 01:43:21 Jamesh: Would you call those predictive models then? 02:05:26 haas: is this forum open for questions? 02:05:45 Gary Meyer: yes! 02:06:39 haas: has anyone touched on untraining 02:07:31 haas: a way to get a model to forget data it was previously trained on? 02:08:37 Peter: lol 02:10:39 haas: or has anyone touched on when to training a model ON your data or when to add knowledge sets. When its best to do which one? 02:13:13 Collard Family: you can instruct to ignore via prompting. also you can remove the embeddings 02:16:28 Jamesh: Once the weights are established, it's like the model has inertia is what is sounds like. 02:24:25 Collard Family: just started learning LC 02:25:13 Collard Family: its a bit more beefy than semantic kernel 02:25:58 Patrick Gronemeyer: https://www.ebay.com/sch/i.html?_from=R40&_nkw=AMD+Radeon+Instinct&_sacat=0 02:26:41 Sean T.: I recommend this tutorial https://github.com/ollama/ollama/blob/main/docs/tutorials/langchainpy.md 02:27:36 haas: thanks patric 02:27:50 Collard Family: 5 minute video w/ soemoen who builds a PDF reader with a q and a front end. https://www.youtube.com/watch?v=u8vQyTzNGVY 02:28:24 Patrick Gronemeyer: https://www.ebay.com/sch/i.html?_from=R40&_nkw=Dell+Nvidia+Tesla&_sacat=0 02:37:44 Patrick Gronemeyer: no 02:42:25 Collard Family: Got to address a family item. Thank you for the presentation! 02:47:04 Jamesh: Too much beer 03:03:47 Patrick Gronemeyer: true 03:04:58 Jamesh: it seems to me that if you left the model alone with its feedback and it was optimizing to accuracy, it would converge towards fewer hallucinations. 03:06:54 Jamesh: Is there a term for this degradation? 03:14:39 Jamesh: https://github.com/queelius/sluug-talk-llm 03:14:51 Alex Towell: ### URL for the Colab notebook on naive n-gram models https://colab.research.google.com/drive/1ak4kOtbIQGXE5kuhhGTd55xu4qRpeZd7?usp=sharing 03:17:02 Patrick Gronemeyer: https://mashable.com/article/samsung-chatgpt-leak-details 03:18:27 Patrick Gronemeyer: https://mashable.com/article/chatgpt-custom-gpt-data-leak-security-flaw-patch 03:18:42 Patrick Gronemeyer: https://www.schneier.com/blog/archives/2023/02/chatgpt-is-ingesting-corporate-secrets.html 03:30:30 Jamesh: Seems like a good strategy would be to have a way to aggregate smaller models into a single intelligence. 03:45:53 tony.c: As many as want to. 04:01:01 Sean T.: https://www.nvidia.com/en-us/ai-on-rtx/chat-with-rtx-generative-ai/ 04:06:11 Sean T.: This was funny: https://thehill.com/business/4476307-air-canada-must-pay-refund-promised-by-ai-chatbot-tribunal-rules/ 04:10:12 Patrick Gronemeyer: https://www.tbsnews.net/tech/ai-teaches-itself-bangla-619070 04:15:37 Jamesh: I've got to run. Thank you so much Alex! Really appreciate it! 04:29:03 tony.c: I want you to act as a Linux terminal. I will type commands and you will reply with what the terminal should show. I want you to only reply with the terminal output inside one unique code block, and nothing else. Do not write explanations. Do not type commands unless I instruct you to do so. When I need to tell you something in English I will do so by putting text inside curly brackets {like this}. My first command is pwd. 04:38:37 Sean T.: Great presentation! 04:38:44 Jamal: Out of curiosity how was this presentation styled based on the readme file. Was there some tool that automated the styles to mimic a powerpoint? Also what OS are you running? I am new to linux and haven't seen that centered bottom app bar before. 04:39:54 Jamal: based on the 'apt' packager manager I guess it's debian based so Ubuntu? 04:40:21 Sean T.: Yeah, it looked like Ubuntu Gnome with some sort of Dash to Panel addon