Jude's AI Blog

Chapter eight provides some prompts to ask ChatGPT and I decided to do this to have a better understanding of Large Language Models. The prompt I decided to inquire about was the ethical concerns with large language models and possible mitigation strategies. The concerns that ChatGPT responded with included bias and misinformation/disinformation. All of these are valid concerns and because of how much AI use has gone up in the past couple of years, some of these issues have happened and continue to happen. On bias, because LLMs are trained on massive text datasets from the internet, they learn and reproduce societal biases related to gender, race, culture, and socioeconomic status. For instance, a model might associate certain professions with one gender or reflect stereotypical viewpoints in generated content. This can perpetuate inequality and discrimination in applications such as hiring tools, content moderation, and automated decision-making systems. Misinformation/disinformation ...

TensorFlow and PyTorch are two leading frameworks that make it possible to build, train, and deploy deep learning models used in many modern technologies. These tools are a big reason why artificial intelligence systems can analyze data, recognize images, process language, and make predictions. For example, TensorFlow is used by companies like Google to power applications such as voice recognition in Google Assistant and image search capabilities in Google Photos. PyTorch, on the other hand, is heavily used in research and experimentation, and Meta relies on it for projects like facial recognition and content moderation. I wanted to know how these frameworks were used in other industries, so I asked ChatGPT. Institutions use these frameworks to detect fraud and forecast market trends, while in healthcare they help doctors interpret medical images and predict patient risks.  For individuals or professionals in data-driven fields like accounting or auditing, learning TensorFlow or P...

I have no idea how Lovable AI is not the most used AI website right now. Before this assignment I had never even heard of Lovable, and it is easily the most impressive website I have tried during these classes. My inspiration for the app I created came from the Pong game on the  Learn Tech Skills  website. I was super impressed with how well the game ran, and how fun it was. Because of this, I prompted Lovable to create a pong app. I linked the final product at the bottom of this blog so you the reader can try it out. While I did have to tweak some features of the app, the result was very similar to what was created off this simple prompt.  The written response from Lovable was this: "I'll  create a retro-styled Pong game with a neon arcade aesthetic - think classic arcade cabinets with glowing paddles and CRT vibes. Features: Two-player local mode (W/S and Arrow keys) Score tracking with win condition Ball physics with angle-based bouncing Start/restart functionalit...

I am glad that this chapter included the Q/A with ChatGPT at the end of it. Sometimes when you are looking at code it becomes overwhelming and hard to take in what is actually happening. The real-world scenario also helped me connect some dots, as I can see how useful reliable predictions are in business. In this example, the tool was being used to predict home values. This tool would be useful for everyone involved in the transaction. ChatGPT was being used to describe the tool, and it even provided ways to allow people to access the tool.  My takeaway from the chapter is to always be looking for problems. Problems are easy to find whether it is in daily life, or the business you work at/own. Once you identify these problems, AI can walk you through solving them. If you and the AI are able to solve your own issue and it was relatively simple, you might even take it a step further and create/post the solution for other people to use. A lot of our problems are shared; the chances ar...

The plan for this week's blog was to try out Replit. I have never used Replit before, so I went to ChatGPT and asked for suggestions on what to do on the website. I was pleasantly surprised by ChatGPT's response, and I thought it gave me some fun but simple ideas to try out. Its suggestions were to make a calculator app, a quiz game, or a Mad Lib generator. I ended up choosing the Mad Lib idea, so I went ahead and took this idea to Replit. I put in my name and email, and that I was a student, chose the free version of the membership, and put in the prompt "I would like to make a Thanksgiving Madlib." Not even 10 minutes later, it created a working interactive app that could generate multiple stories.  Okay, maybe it was not completely finished after ten minutes. I started inputting different words, and in the middle of this doing the website refreshed, and Replit replied with this explanation.  " Oops, it looks like the Mad Libs form is still a bit shy! When you ...

I have always wondered how our phones can guess the next word or phrase we are going to type. Natural Language Processing (NLP) seems to be the answer to that, as it is the reason that computers can actually understand and work with human language. I am still impressed with the process, because it is either accurate or very close to predicting the rest of our sentences. I believe iPhones even adapt to the person who is using the phone and begin to recognize the owner of the phone's way of talking/lingo. Learning about text classification, sentiment analysis, and language modeling helped me see how AI can make sense of huge amounts of text data.  Using NLP to sort emails or analyze social media posts is a good example of how these models work and are used in real life. I also found it helpful to learn about important steps like tokenization, padding, and embedding, which prepare language data for machine learning. It makes sense that models like ChatGPT are built using these same te...