{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Crazy Wisdom","title":"Where do you fit in when it comes to the excitement vs fear spectrum when thinking about AI? - Matt Bunday","html":"","width":"100%","height":180,"duration":3300,"description":"Matt Bunday works in crypto. He loves to rock climb, martial arts, and think about underground psychedelic therapy\n \n What is the biggest problem you’ve faced with knowledge management at your companies?\n Slack \n Dropped balls in communication\n LLMs might be the response\n Why do you think slack spread so fast even though its not the best product?\n It was a step up from email\n IRC was a component\n More friendly for non-engineering\n What do you think about the complexity of slack?\n Twist is an alternative\n Why is information architecture such a challenge?\n How do LLMs fit into this?\n What would happen if slack created a LLM or plugged one in?\n Slack workspace plug in\n What is a retrieval plug in?\n Universal adapter for any type of data\n Who are the incumbents in the slack knowledge management space?\n Guru startup\n What is the difference between information management and knowledge management?\n Knowledge management is a higher level synthesis\n Information management is siloing related types of information\n Data types \n Group related to information around people\n What are your thoughts on the membrane?\n Siloes\n Privacy is where LLMs can be very innovative\n If we were to share a LLM\n We can both specify a privacy policy to the LLM and it will follow it\n LLMs can intuit the privacy public distinction\n Are you using LLMs to code for you?\n Copilot\n Issues with difficulty to prompt it correctly\n You had to write comments to prompt it\n Inline suggestions were not good\n Is it better now?\n Haven’t noticed a dramatic improvement\n Hard to prompt it to code in certain styles now\n GPT 4 is way better for a starting off point for projects\n Helpful for conversion processes\n What are the things that GPT4 has not been helpful for you?\n You have to chunk it\n What about building systems with GPT4?\n Code completion cool called tap9\n Train the model against your local code\n What are some other things about KM that we can use tools for?\n Shared LLM for the family\n...","thumbnail_url":"https://img.transistorcdn.com/UZbrDrlO5VTfDNcq188THwbv0T09vcmLyzx3BcPI9bs/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS81Y2Rj/OGFiMTYyMGFkNTM5/N2NjOWI2MWM5YzQ1/YTc2Ny5qcGc.webp","thumbnail_width":300,"thumbnail_height":300}