I plead the fifth.
TLDR.
I have some limited local AI training and inferrence shit set up using pytorch, transformers, hugging face, etc etc.. Check out https://lmstudio.ai/ for local AI easily without all the nerdy tech shit
You are viewing a single comment's thread from:
I want to adapt AI to plan aquaponics systems per customer requirements, to provide lists of parts and how to employ them to attain specified nutrition per local conditions. I was hoping you could achieve a promptable model that could be copied and deployed to people setting up aquaponics systems. I have no experience with using AI, and certainly not using it for this specific purpose.
Thanks for piping up at all!
Probably don't need an AI for that.. Just a formula / algorythm.
tl;dr* below
Well, the intention is to be able to take the environment proposed (the circumstances of the intended user) and design a system for them, per their specific goals (they may love carrots, or hate them, etc.). I don't see how an algorithm can handle the 3D volume of indoor space available, angle of the sun through extant windows, menu planning, and the full panoply of potential components/species that can comprise such system. AI is a large library of algorithms, which seems to me necessary for dealing with all the many aspects of making systems with requirements of expertise in farming, hydroponics, animal husbandry, and etc., that come into play - and eliminating that need for expertise in all those fields. While I have no experience in prompting AI, nor in training AI, I am aware that providing an appropriate training dataset can enable privately owned AI to generate useful responses to natural language prompts. I am not well apprised of your specific experience, but am aware you certainly have general acquaintance with these requirements, better than anyone else I know or follow.
Additionally, I intend to wed a variety of table top production technologies into holistic systems people can adopt that replace their dependence on centralized production with independent production. I expect each of the technologies will require similar management by AI in order to integrate the wide variety of production technologies available into a system designed for the specific circumstances and goals of the user.
This is my goal, at least. Clearly, while decentralized means of production of almost any good or service underlying the blessings of civilization on the market today exist, almost no one can deploy all of them, so the need is to enable people to adopt those that make sense in their circumstances and to automate to the degree possible management of those systems so that, to the extent possible, the transition to independent production is seamless to the consumer in terms of their access and use of products. This also will require an ability to trade and exchange for goods and services with peers that further enables people to specialize in market production for those products they are particularly well situated, or apt at, to produce, again, likely requiring tracking of many producers and trade potential, etc., that AI is well suited for.
The goal is to enable free and independent people to themselves be the technocrats that govern them, to prevent overlords from rendering a class of thralls to subjugation by controlling their access to essential goods and services. Kissinger pointed out that an institution that controlled oil controlled nations, and one that controlled food controlled populations. If populations themselves independently provide all their necessities and the goods and services that create the blessings of civilization they control their freedom and felicity, and that is my ultimate goal. Because of the primary nature of food and the simplicity of using a barrel of catfish to feed a family, yet the potential to automate and extend production is relatively facile, it seems to me that starting with aquaponics would be a good idea.
*tl;dr Perhaps you could suggest a way to efficiently get up to speed in acquiring personally owned AI, training for specific purposes of designing aquaponics systems, and designing prompts to generate responsive aquaponics designs. Based on your experience with AI your advice in this regard would be very useful to me, who has none.
AI is good at making fluff / bloated speech, but even the best available LLM's these days still do not actually even remotely understand the concept proposed by words in any given field.. they are simply logic networks that calculate what the most human / normal response is in a string one word after the other.
What you're looking to build here would not really benefit from having a large langauge model directly talking to customers.. Mainly because it doesn't really add value. In the case of retail like stated above all of the metrics, conditions and variables used to determine the best package for their budget should be a relatively simple algorithm/ function and/or formula.
A sale package template that is auto populated given the user defined metrics and whatnot is likely the way to go, Adding AI to write fluff pieces about the setup is fine I guess but as far as AI goes you're not really going to gain anything trying to train an AI to do what should be done usng "back of the napkin" style maths.
This shit isn't a magic answer to business nor development.. Perhaps a strictly narrow band sales agent with a few AI elements in it could streamline your customer service and or salesteams work.. But as for having a model trained on aqauponics would be pretty redundant.. given that AUI doesn't even know what the fuck water or platns are, let alone how things like light or nitrogen levels effext yield.
If you want a personal AI, Start looking at huggingface, pytorch, LM Studio, github, etv etc to start learning to build and or import models and training data in order to start experimenting.
https://huggingface.co/
The site above has tons of publicly available models, and with a small amount of python or javascript (both sometimes) you can easily explore, incorporate and get your feet wet with AI stuff at home.
Off topic: I've been playing with the idea of a tilapia & vegetable aeroponcs/aqauponics system for a while. Love the idea of that shit.
I very much appreciate your substantive reply. I am sad AI is so dedicated to language the algorithms are incapable of being trained on other metrics, like sorting bolts and nuts. I don't want to create sales pitches. I want to sort nuts and bolts, to take available square footage and watt hours of light available through a window and determine how many of what plants can be grown there and how many crawdads or tilapia it will take to fertilize them.
Thank you very much!