World of Text: Smart Home Automations You Won't Find Anywhere Else

[@aidanb8719]

This is an amazing idea and has so much potential, it would be like having a best friend to help you along in life.

[@TheOfficialOriginalChad]

Man this channel is so so so good! What a phenomenal intro

[@technithusiast]

Glad you liked it!!

[@michaelolsberg2458]

I really hope you continue to explore this more. I am very intrigued by it, and after watching this video, I think I really want to look into it more

[@technithusiast]

The concept is very fascinating! Here's part 2 exploring the concept: kzbin.info/www/bejne/d3yoaohqZ7CMosU

[@Mephmt]

This is exactly what I've been dreaming about since I first learned about LLMs. My first attempt was trying to figure out how to make a custom GPT call the Home Assistant API.

[@dankkster]

Great video. This definitely has a lot of potential. One thing you should probably add to your "world of text" is a warning about the use of openai as opposed to a locally sourced llm. Open AI irreversibly creates a profile and stores, every question, every note, every piece of data you feed it. Not only do they build future models with it, but they store that data indefinitely and already secure the rights to do pretty much whatever they want with your data which is now their data. Some people may not care. But it would be good to at least let them know of the consequences of using open AI for something like this.

[@JB-fh1bb]

@1:01 Caution: it's not specifically text that's generated, but we can describe most of it in text. The caution part is that LLMs are great at text now but if you absolutely lovk in to text you'll miss out on the evolutions of AI. They're already starting to understand audio for example. Imagine if you transcribed your conversations in to text then later asked an AI "when are times that my friend feels joy?" WOT wouldn't have access to the intonation and therefore would struggle to answer that question (and then how could it recommend the perfect gift?)

[@georgehooper429]

Well done making me consider something I never new I needed. Now I want one! TBH I did not watch your Notion database video. I thought to my self meh, why… But now I think I need to watch it to catch up. Well done (part two)! I have to be that guy for a moment. I’m thinking of the legal ramifications of recording and transcribing conversations into text for future recall. The laws in the US vary from state to state, but recording and transcribing dialog with a second party especially in a two party state would be problematic. Would everyone needs to sign a release so you could take notes and recall what someone said and when? I know this hits on the can we/should we rule, but you know if it exists someone will abuse it. While your project is only to prove it can be done its something that will need to be thought about later. While Ollama, or matter of fact current hardware technology isn’t there yet all of this data could be fed into your personal self trained AI model. You can take text documents (like MS Word) and feed that into a personal model. You would almost need to start with a standard LLM that has access to the information on the internet and then augment that data by having it learn about your life and experiences. I think its an interesting concept. By the way good job on using Generative AI to assist building your stream thumbnails. It adds kind of an artsy flair to the icons.

[@michaelduncan8345]

So... the smart home enthusiast's Pensieve.

[@rkaid7]

I’m 15yrs deep into what ud call personalised crm. But Ur nibbling around the goal. It’s AI Assistants, always has been. Not just text. Humane Pin / Meta Ray Bans with camera. Feed it everything. Everything RAG’d, it working on 100’s of projects ur both working on. Interrupting u when milestones are met. Full root of an operating environment, eventually full control of the systems running it.

[@InspiredInsights4U]

I think the project would end up being too big for just an ocean database. Probably what you would end up doing is having more robust database built on a cloud server than when you’re working and recording and doing other things with all this information you’re gonna have to process it and that’s where an AI automation Would come where you would feed multiple aspects of the information through the automation allowing it to like you said summarize identify bullet points but you don’t have to do a lot more than just that you’re gonna have to build all of this and then have it fall into individual that could be About your thoughts and your individual information during the processing of the stack you could look for the relevance whether it’s something to world events or whatever and have that recorded along with it then years later as you’re trying to track through the process, you would have all the information that you would need Interesting, but I think there’s a lot I don’t think would be powerful enough to handle it my opinion

[@Renegadegit]

😊