2/14/24. AI and New Possibilities
Frank L -
First of all, just to clarify, I’m not working with sentient AI or anything quite on the level of the large language models like Google’s Gemini or OpenAI’s ChatGPT. The definition of AI is pretty vague in general, so basically any program or software that has the ability to perform complex tasks like analyzing, reasoning, and learning can be classified as artificial intelligence. Most programs only have the ability to carry out specifically written instructions, but AI-powered programs have basic learning abilities and can train themselves to improve performance and efficiency by gaining some experience.
There are two main approaches to AI, machine learning and deep learning. Machine learning uses statistical algorithms to process mass amounts of data for the purpose of finding patterns or making predictions. Deep learning is a specific type of machine learning technique that is modeled after the way a human brain works. These programs are structured with complex neural network architectures that process data through many different layers of nodes that work together kind of like neurons firing in a brain. It sounds cool, but most current models are extremely inefficient and require insane amounts of computational power. So yeah, that’s why so much research still needs to be done.
Anyways, back to my senior project, I’m currently helping out as a student assistant in a research group at ASU that currently focuses on the practical applications of AI systems in digital manufacturing. I’ve actually been doing this for a while now. I first got introduced to the professor over the summer last year and since then, I’ve helped out with tasks like writing literature reviews on research articles relevant to the projects, analyzing some of the python programs, creating images from raw numbers and data, and sorting out the datasets. The actual AI programs are way too high level for me to understand right now, so I’ve mostly been learning things along the way and just trying to reduce the workload on the PhD students that I’m helping.
Recently, I’ve been aiding a project on generative adversarial networks (GANs) for the manufacturing process of resistance spot welding. I just got handed a subtask on identifying cracks in steel from surface scan images, so I’m still trying to figure that out. In my next update, I’ll give an explanation on how a GAN works and hopefully have some more concrete details on my current tasks.
Comments:
All viewpoints are welcome but profane, threatening, disrespectful, or harassing comments will not be tolerated and are subject to moderation up to, and including, full deletion.