Error converting content: marked is not a function
- Would you trust me, Tate, the same if I was obese? - @Andrew Tate - THINK about it. - How could I promise to help you escape the Matrix if I couldn’t escape from a Hamburger? - I am so pumped and inspired to build my own P/AI Compute Framework - creating this right away :) - Inspired as I watched a Tech/Langchain screencast. Which got to to see Tech/Langflow - graph vizualization! something I always wanted at Cogito. - Things I guess my compute can do - Multi model - text, image, video - element type :) - Real time?? That's a biggie. Langchain and others are all thinking batch - Will start with python I gues - could then get this abstraction for js and rust - 11 PM - #practice/hacknight init. #TrialAndError.AI - The implementation of Agents is inspired by two papers: the [MRKL Systems paper](https://arxiv.org/abs/2205.00445) (pronounced ‘miracle’ 😉) and [the ReAct paper](https://arxiv.org/abs/2210.03629). - 4 AM - Personal Agent Idea #P/AGI Agent - Mechanics and info flow - Seed - Who am I? A personality. An Ego! Ego generated action. For the agent perspective we have an input. But you don't get caught by it. The input hooks you and bring ego to life. So say the Self - which is all observing - creates a person! - Person has personality. Memory. A sort of learning as well. **Always changing.** It's writing it story as it run and completes tasks and outcomes. Maybe even version it's own evolving personality. - Hm - For Person to have personality and a story - it needs **Time and Space** - ?? time could be age. Matching with real time or total compute time. It won't make sense without being relative to something. - For space - no clue right now - Person - like GTD - Imagines the outcome #boom. This kicks offs actions to achieve the outcome. - Person (agent), has a plethora of tools to accomplish each action. - Persona may have a learned memory and understanding of each action. Including what he is really good at and what sucks -- here sucks would be lack of confidence in result or time it takes or just not sure if he has the right tool for the right job. - Thus, Before taking any action it spits out a confidence score. Then it tallies and update this score after the action is completed and result observed. The audit trial of this score is updated in memory. - This can be a RELATIONAL DATABASE! Duh. - Person also adds/updates list of tool he wishes to learn and excel at. Learning from tasks. - Person agnet can also talk to other agents. They would be different from tools in the sense of real world. While person can get better at existing tool, it cannot control the agent. It can definitely replace or use different tool or agent. - Preference would be to use the tool. Or have a set of trusted agents who are good with set of tool. This would be Persons' way of managing complexity. oooo. - Example - Coding Agent - for all things coding. It maybe coding tools. Shared tools. - Every interaction is an opportunity for reflection. To improve and learn. **To get to it's final goal of learning all tools and to be able to complete all tasks.** - Meanwhile Self is watching. Just the witness. It contains the entire process. Only Self knows the Self. - Some other ideas to explore - Person leisure time? Maybe meditation and discussion with therapy agent. - A routine audit of it's own capabilbity, improvements, and personal goals!!