Learning to program AI

[BDR99]

Hi,

I have a background in electrical engineering and trying to break into the AI industry. trying to learn without guidance is really overwhelming (like RNN, Word2Vec, Adam Optimizer, Cross-entropy loss, just to name a few ).
My main concerns are 1) how to program the model since that's critical to how it responses and "train to think". 2) how to program the datasets.
Are there any suggestions on where to start as a project? any pointers? what simple github projects I could do in the beginning as a stepping stone to gain experience in programming the AI model and datasets?

At the end I would like to try to make my own AI program and maybe turn it into a Saas platform.

right now I have been learning PyTorch as I read that its is one of the prominent framework for AI and Machine Learning.

Thank you!

BDR

 

[peteros]

99% of "AI startups" just uses big providers (like OPEN AI), and just build some wrapper on top of them.
I don't think there is many people here, who actually create some ML or LLM software. For sure not me.

And I don't even know if you can actually do something like this by yourself (in terms of something that will have any value to any sort of customer of course, I'm not talking about equivalent of to-do app here...).

Sad, but this is my opinion.

 

[BDR99]

I appreciate your response to my thread.

How extensive is your experience with AI programming? Because I’d like to pick your brain some more.

I am not aiming to create an exact clone of ChatGPT like OpenAI's, but rather to train and curate AI for specific medical applications based on specific documents, X-ray, for, diagnostics, early detections, etc. I am open to using OpenAI, Claude, or LLaMA as the AI engine for the platform.

So are you saying these platforms don’t have anything special? They are just chatgpt on a wrapper? Because I’d like to create something similar like these…

• https://www.documind.chat/
• https://mindgrasp.ai/
• moltar.ai

On a separate note, the models on HuggingFace are used to train AI, correct? How do they accomplish this? Because this is what I think I need to learn.

Models - Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co

 

[Kristy_Poole]

1. Check out these resources for project ideas:
   • PyTorch Examples on GitHub
   • fastai Library
2. Learn and Practice:
   • Kaggle Competitions: Great for hands-on practice.
   • Online Courses: Try Coursera’s Deep Learning Specialization by Andrew Ng.
3. Long-Term Goal:
   • Identify a problem your AI can solve, build a prototype, gather feedback, and refine your solution to eventually develop a SaaS platform.

For more practical insights, refer to these sources:

• PyTorch Official Documentation
• Kaggle Learn

 

[BDR99]

thanks guys. these advises help.

what do you think of Grok then? what kind of frame work is this? how do one use this?

GitHub - xai-org/grok-1: Grok open release

Grok open release. Contribute to xai-org/grok-1 development by creating an account on GitHub.

github.com

 

[peteros]

• https://www.documind.chat/
• https://mindgrasp.ai/
• moltar.ai

First one can read PDFs, so looks like something more than GPT. That's surely some OCR for reading content and then GPT for 'interactions'. So basically yes, some sort of wrappers.

From what you said, you want to train your AI with data - I don't know about others, but for sure you can do that with AWS. But I think others likely give such feature as well.