Program AI yourself?
Hello there,
A side question that has been on my mind. An AI is created/programmed by humans. But can an amateur with little to do with IT simply create an AI like that? Let's say, for example, something similar to Chat gpt. 🙄 I'd like to know how an AI works, let alone whether you can create an AI "just by guessing." I think there are massively powerful servers behind it. It probably won't be a 0.8, 15, or 15-inch home office PC or gaming PC. I'm thinking more in the direction of "high-performance computers" that are the brains of the AI.
Who knows how AI works?
Who knows if you can create your own AI?
Thank you
(3 votes)
Loading...
I’ve defined for myself: As long as I’ve never had one MLP-Algorithmus itself programmed and used, I’m too stupid to know what AI is. I have developed and used learning algorithms for half my working life, but that was not AI, but regression analysis to determine the coefficients of mathematical models (empirical equations) from measurement data.
To program algorithms that you can see as AI for learning purposes, of course, you don’t need a giant server that goes on normal PCs, because you just want to playfully and modelly understand how it works and not make ChatGPT competition.
If I ever give up the motivation for this – I don’t have to do anything anymore, am a pensioner – then I will try to predict with MLP lottery numbers for learning and training purposes, because the algorithm will fail, and it is definitely particularly interesting and educational to see how it fails to imagine the limits of its possibilities.
There are thousands of man-years of development in it by highly educated people and data centers whose size is measured only in football fields.
It’s not gonna be “self-programming”. Not in a single human life.
The topic “KI” sounds rather complicated. Above all, an AI learns by “question / search”
“more complicated” is a really cute euphemism.
Yes either a RAG application or fine tuning
Are there special programs? Or is it more (which I can imagine) about writing code lines, off?
The Retrieval component of the RAG application is logically something you need to do yourself, which is then executed on each request
For finetuning there are some libraries but I believe you also have to write a little Python code, but it is only executed once
I actually referred to LLMs and not to these assistants who actually only control different models and algorithms
I got two Alexa home. I know it’s an AI that learns the question. So they know what you’re interested or what you’re looking for. 🙄 But apparently the subject of “own AI” would be more difficult than you think. Doesn’t sound easy. Have it ever presented to me so easily 🙄