/robowaifu/ - DIY Robot Wives

Advancing robotics to a point where anime catgrill meidos in tiny miniskirts are a reality.

Canary has been updated.

Build Back Better

Sorry for the delays in the BBB plan. An update will be issued in the thread soon. -r

Max message length: 6144

Drag files to upload or
click here to select them

Maximum 5 files / Maximum size: 20.00 MB


(used to delete files and postings)

Have a nice day, Anon!

Open file (8.45 MB 2000x2811 Chii_w_Atashi_kawaii.png)
Open file (442.93 KB 3109x3840 percept_UML.dia.png)
Cognitive Architecture : WIP Kiwi, NoidoDev 09/17/2023 (Sun) 20:38:37 No.25368
"An algorithm without a mind cannot have a heart." Chii Cogito Ergo Chii Chii thinks, therefore Chii is. --- # Introduction With cognitive architecture we mean software that ties different elements of other software together to create a system that can perform tasks that one AI model based on deep learning would not be able to do, or only with downsides in areas like alignement, speed or efficiency. We study the building blocks of the brain and the human mind by various definitions, with the goal to create something that thinks in a relatively similar way to a human being. Let's start with the three main aspects of mind: * Sentience: Ability to experience sensations and feelings. Her sensors communicate states to her. She senses your hand holding hers and can react. Feelings, having emotions. Her hand being held bring her happiness. This builds on her capacity for subjective experience, related to qualia. * Self-awareness: Capacity to differentiate the self from external actors and objects. When presented with a mirror, echo, or other self referential sensory input is recognized as the self. She sees herself in your eyes reflection and recognizes that is her, that she is being held by you. * Sapience: Perception of knowledge. Linking concepts and meanings. Able to discern correlations congruent with having wisdom. She sees you collapse into your chair. She infers your state of exhaustion and brings you something to drink. These building blocks integrate and allow her to be. She doesn't just feel, she has qualia. She doesn't see her reflection, she sees herself reflected, she acknowledges her own existence. She doesn't just find relevant data, she works with concepts and integrates her feelings and personality when forming a response. Cognition, subjective thought reliant on a conscious separation of the self and external reality that integrates knowledge of the latter. A state beyond current AI, a true intellect. This thread is dedicated to all the steps on the long journey towards a waifu that truly thinks and feels. --- # Rules - Keep your comments short; lengthy speculations aren't helpful - This is about implementation; advance only useful perspectives, please # Topics we don't want to discuss in this thread - The question if we should even work on such things as AI or cognition - Beliefs stating that we can't do what we want to do - Esoteric discussions about "sentience" and "conscience" - Suggestions about only using deep learning and similar techniques - Suggestions that we should put all of it into one deep learning model - Suggestions that we should not use censored or mainstream models at all - Debates in philosophical lingo - Comments from people who don't even watch videos on AI or only on deep learning and tutorials, or only about warnings in regards to the dangers of AI. - Wild ideas on how to do things totally different but aren't useful, there are other threads for this: # Related threads - try to use them if something fits - The old chatbot thread >>22 is now the one for cognitive architecture >>24783 - The next GPT thread which is currently >>250 could be one about the tech of LLMs, and only that. Or it also goes into >>24783 - Or we make one for anything related to language responses, so scripted chatbots and LLM, minus the more elaborate thinking - Abstract debates about conscience and sentience should go into the thread about "philosophy" >>11102 - Fundamental doubts and such also. It is also about the postings which would only be disruptive in the "cognitive architecture" thread. Basically the things which the people implementing it don't want to be bothered with. - Morals and such >>17125 might also have an abstract implementation problem. - We have threads on all kinds of personality and psychology. >>77 >>250 >>18 >>11102 >>17125 >>16217 >>15047 >>2731 >>22 >>111 >>107 >>152 ... and more in the catalog.
Edited last time by Chobitsu on 09/18/2023 (Mon) 00:53:46.
# Cognitive Architecture - Separating it from conversational AI is probably impossible, since we need to use it for inner dialog and retrieval of data from LLMs - I'm not sure if optimizations for conversations, like faster understanding and anticipation where a conversation goes, should also go in here or in another thread - We need to approach this from different directions: Lists of capabilities humans have, concrete scenarios and how to solve them, configuration optimizations - The system will need to be as flexible as possible, making it possible to add new features everywhere # Definitions of conscience which we might be able to implement - ANY comment on conscience has to make clear to which definition it refers or make the case for a new one which can be implemented in code (including deep learning) - <list, it's five or so> # List of projects working on Cognitive Architecture or elements of it - Dave Shapiro: https://www.youtube.com/@DavidShapiroAutomator - OpenCog - Some guys from IBM Watson worked on it - LangChain - - ... <there are more, we need a list> # List we need >>22488 # Related technologies - Foundational management of the parts: Like Supabase? - Firebase alternatives: https://youtu.be/SXmYUalHyYk[Embed] - GraphQL frontend for sqlite https://youtu.be/KfE1Tm1gZUU[Embed] - Graph databases and modelling of graphs https://youtu.be/OzDG68VvPxY[Embed] - Traditional Parsing - Vector databases: https://youtu.be/klTvEwg3oJ4[Embed] - RDF: https://en.wikipedia.org/wiki/Resource_Description_Framework # Links to repositories - Dgraph: https://github.com/dgraph-io/dgraph # Elements - Long and short-term memory - World model databases - Filters for actions and responses based on configuration # Tasks - Going through AI related threads and making an overview with links - List we need >>22488
Edited last time by Chobitsu on 09/17/2023 (Sun) 20:54:52.

Report/Delete/Moderation Forms