Improve ChatGPT: Modular, Adaptive, Smart LLM | Inside ChatGPT

Key Takeaways

hi Community you ask me can we build a smart bot or I say a generative monolithic AI versus a self-configurating smart AI look at this generating functional of a vector space you see what I mean look now in comparison to a monolid and if it peers through the surface of this monolith we have more or less quantum foam now what is the difference I tell you not the filaments a neural network conic terms here we learn Finance here we learned science here the system learns all about policy so why does monolithic structure why only on Microsoft cloud because it includes everything Bloom for example ran 117 days in France and did cost about 3 million so it's an alternative in Microsoft For What Microsoft cloud is for everybody so let's have a detailed look at them not at a monolithic system but at a modular system this little Cube on top is my neural infosystem I built a sentence Transformer system for semantic search only on Quantum field Theory in this video I show you how you can do it my data I have in a Delta Lake structure here's the video for you to copy I have there about 12 000 Publications in every week I download new one every month I fine-tune it and twice a year I update the whole system with a free training imagine I could reconfigure the system to include biology chemistry currently I can't imagine the system would self-adjust based on my prompt it includes toxicology or genetics now if I have a system that is modular I could exchange part of it I could update the port and bring it back in 26 percent of you want more mathematics here we go a homework for you combine the lead group architecture with the topological massing passing on simplicial complexes and for the rest of my viewers here look this is the problem this cube is my expert information retrieval system on a neurological level but you see the border the active border is the problem not only in the information flow but it exchanges the information about the architecture of the system itself all the elements within this Cube have to reconfigure based on the task given to the system so I want a modular system I want the system to be able to open the transfer learning so every month I can re-learn it on actual data I wanted to self-adjust and I wanted to have a local smart response of the system I have Smart hops here in my home why not use them this is the visualization I found for my solution smart modular self-configurating adapted to the task