Working With Walmart, Exploring AGI, and Unlocking my Artistic Potential With GANs
Vinaya Sharma | December Newsletter โ๏ธ๐
Hey! If youโre new here, Iโm Vinaya and welcome to Bits and Bytes! Iโm a 16 y/o emerging technology enthusiast hoping to change the world! ๐
In this monthly newsletter, I share bits and bytes of my learnings to help you grow from my key insights! ๐ฑ Stay tuned for bits of self-improvement techniques, and bytes of groundbreaking tech innovations.
December has been all about building, learning, and reflecting. Check it out!๐
Working With Walmart to Improve the Customer Experience ๐
As I mentioned in my last newsletter, during November and December I was working alongside my challenge team at TKS, and digging deep into how Walmart can improve the customer experience. Specifically, Walmart tasked us with developingย simple & scalableย in-store digital touchpoints or experiences that further connect customers to our digital journey. With the 21st century revolutionizing nearly every industry it is important that Walmart innovate to remain competitive. ๐
After loads of researching, brainstorming, prototyping and validating my team and I boiled down Walmart's greatest opportunities in terms of improving the customer experience while capitalizing on missed revenue and set out on a mission to improve the current system.
As promised hereโs a breakdown of our ideas to improve the shopping experience:
In Walmartโs current customer journey they fail to provide a personalized experience as compared to its online competitors. Due to the overwhelming product options and the lack of a personalized experience, in-store shopping is undesirable for customers. As a result, Walmart loses between $36B and $60B in potential revenue per year to eCommerce competitors.
In order to combat this issue my team and I proposed the integration of Frontend and backend digital touch points throughout the customerโs journey to create a seamless in-store experience with personalized incentives, time-efficient visits, and an emphasis on community engagement.
We hope to transform Walmarts current customer journey into a digitalized one, with various enhancements that will save time, and money while improving customer retention.
Our 3 recommendations comprised of an improved app, an interactive recipe shelf, and a queue management software
Customers can โStart Their In-Store Journeyโ through the app to shop under an โomni-channel profileโ for a personalized shopping experience through real-time recommendations, reminders, and individual incentives based on aisle location and dwell time.
SparkYourRecipe, an interactive recipe shelf integrated with a new community engagement tab on Walmartโs app. A physical display of weekly popular in-store ingredients and uniquely curated recipes for customers to try and be inspired by. Customers receive rewards when sharing their dishes, reviews and memories on the app, allowing them to see their posts in store, and increasing engagement within the Walmart community. ๐
LineBuster. Quantifying real-time store footfall data and queue demand to predict cashiers needed up to 15 to 30 minutes ahead of time. Ensures checkouts are running efficiently by reallocating employees where needed to reach Walmartsโ wait time quotas of no more than 3 customers in line at a time. โฐ
Huge thank you to Lynn Jennifer Kulli, Najeeb Khan and Sophia Story for meeting with us in order to brainstorm, refine and validate our ideas. Your input and advice coming from your experience of working in your respective industries were invaluable.
And finally, I want to take a moment to give a huge shoutout to my team Priyal, Tanvi, Zara and Krish that made every minute of working on this project worth it. I wouldnโt want to spend 5 am to 3 am work sessions with anyone else!
Definitely check out our standalone deck to learn more about each of the proposed solutions. We will be hearing back from the Walmart Blue Labs team early in the new year to hear which recommendations have been selected for review and potential implementation. I canโt wait for the results and I hope my team gets to present our ideas in more depth to Walmart Executives!
Artifical General Inteligence Sprint
Imagine a machine that can do anything that a human can do โ think, learn, problem-solve, and even create. That machine is what we call Artificial General Intelligence (AGI). AGI represents a major leap forward in the field of artificial intelligence, and its potential is truly mind-blowing. ๐คธโโ๏ธ๐คนโโ๏ธ
With AGI, we could potentially solve complex problems that are beyond the capabilities of current AI systems. It could help us develop new technologies, find solutions to global challenges like climate change, and even revolutionize entire industries. The possibilities are endless!
But achieving AGI is no easy task. It will require a significant amount of research and development, and it is not yet clear when or if it will be possible. Nonetheless, researchers and developers are working tirelessly towards this goal, driven by the belief that AGI could have a huge impact on the world.
Early this month and I completed an AGI sprint to better understand the current state of AGI, the proposed methods of achieving generalized intelligence and the safety and ethical implications of such technology. Check out our Memo. ๐
Building Generative AI
Generative AI is a type of artificial intelligence that can generate new content or data, from artwork and music to ideas and concepts. One of the most powerful tools for generating new data is the Generative Adversarial Network or GAN. A GAN consists of two neural networks that work together to create new data that is similar to a training dataset. The back-and-forth between the two networks helps the generator create more and more realistic data. ๐ก
The potential of generative AI and GANs is truly exciting. They could revolutionize the way we create and consume media, improve data analysis and decision-making, and even generate new scientific discoveries.
The future of generative AI is an exciting one, and thatโs why this month I spent a lot of my time understanding this technology and building projects. I built the same iconic project using four GAN architectures in order to understand the differences and benefits each provides. Of course, I worked with the MNIST handwritten digits dataset and I specifically build out the Vanilla GAN, Deep Convolutional GAN, Wasserstein GAN, and Conditional GAN.
You can find my results of the CGAN below. Although it took a lot of training the final results are insane, just by requesting a digit the model is able to return a unique drawing that perfectly resembles the training data. โ๏ธ
I Made an AI Turn me Into an Artist ๐จ
Finally, all of these GAN implementations came together as I embarked on a journey to turn myself into an artist.
As a self-proclaimed artist, Iโve always dreamed of creating beautiful, breathtaking works of art. But sadly, my talent seems to have been left behind in the womb. My stick figures look more like spaghetti and meatballs, and my portraits resemble the love child of a potato and an octopus. ๐ฅ ๐
But I refuse to let my lack of artistic ability hold me back. Thatโs why I decided to take matters into my own hands and create a GAN (generative adversarial network) to turn my nasty sketches into beautiful photographs.
Using a combination of machine learning algorithms and a dataset of high-quality photographs, I built a GAN to transform my scribbles into realistic images that even Monet would be proud of. Itโs like magic, except I had to put in hours of coding and debugging to make it happen. ๐ฉโ๐ป
This project really tested my understanding of working with deep learning and PyTorch as there was no tutorial available to lead me on a project in this domain. I had to rely mainly on the CycleGAN research paper and stack overflow to figure it out, but I am glad I did and learned so much in the process.
You can check out my article, git repo, and youtube video for this project to learn more about how I built my Sketch to Photograph generator and view some results with me!
Be sure to subscribe and follow in order to stay updated with my latest ML projects. Also feel free to reach out if you have any questions, comments or concerns regarding the project and Iโd love to help out anyone wanting to build with AI!
Bytes of Brain Food ๐
Each month I consume loads of content through articles, podcasts, and videos. Here are my favourites featuring several technological breakthroughs. ๐ฅ
OpenAIโs Point E: Point E is a method for 3D object generation that produces 3D models in just 1-2 minutes on a single GPU. This is a significant improvement over the state-of-the-art methods, which typically require multiple GPU hours to generate a single sample. Point E works by first generating a single synthetic view using a text-to-image diffusion model, and then producing a 3D point cloud using a second diffusion model that conditions on the generated image. While this method may not produce samples of the same quality as the state-of-the-art, it offers a practical trade-off for some use cases due to its significantly faster generation time. The authors of the paper have also released pre-trained point cloud diffusion models, as well as evaluation code and models for others to use and build upon. ๐ช
Google Brainโs CALM: CALM (Confident Adaptive Language Modeling) is a new method for improving the efficiency of language models (LMs) during text generation. Many recent breakthroughs in natural language processing have been driven by LMs like T5, LaMDA, GPT-3, and PaLM, which have demonstrated impressive performance on various language tasks. While increasing the size of these models has been shown to be important for improving performance, large LMs can be slow and costly to use in practice. CALM addresses this issue by selectively allocating more computational resources to the harder prediction tasks, allowing the model to generate text faster while maintaining output quality. This approach is motivated by the observation that some next-word predictions are easier than others, and current LMs devote the same amount of computing power to all predictions regardless of difficulty. CALM was presented at NeurIPS 2022 and offers a promising solution for improving the practicality of LMs in real-world applications. ๐ค๐ค
Chat GPT: Of course, this monthโs tech breakthroughs would not be complete without talking about Chat GPT. OpenAI released ChatGPT, an artificial intelligence (AI) chatbot system that allows users to ask questions using natural language and receive conversational responses at the end of November. The bot is able to remember a thread of conversation and use previous questions and answers to inform its next responses, drawing from vast amounts of information found on the internet. While ChatGPT is able to provide useful answers in areas with good training data, it is not yet omniscient or smart enough to replace humans. OpenAI warns that ChatGPT "may occasionally generate incorrect or misleading information." ChatGPT is able to recognize patterns in text and deliver more useful dialogue with human assistance, but its answers may not always be accurate. ChatGPT is intended to showcase the capabilities of a large and powerful AI system, and more than one million people have tried it out since its launch. Throughout the month of December startups and established companies have all been inspired to use Open AIs open source APIs and level up their own chatbot systems. We can now see fantastic improvements and implementations of language models with Jasper.AI, Quoraโs Poe and much more! ๐ฌ
End of the Year Reflection ๐ฑ
Each month I typically talk about the mindsets Iโve been practicing in order to propel my self-growth, and share techniques I have been applying to help you learn from my growth and failures. Considering it is new years eve and the end of 2022 I thought Iโd change it up and reflect on this year as a whole and set some intentions before heading into the new year.
As we kick off the new year, it's the perfect time to set some powerful intentions for the months ahead. I'm not the type to be obsessed with new year's resolutions, but I do see this time of year as an opportunity to make improvements. During the holiday season, I tend to have the most motivation and energy, and I want to make the most of it.
One of my top goals for the new year is to boost my time management skills and build some seriously impressive AI projects. To make this happen, I'm going to follow through on my time-blocked schedule by training the mindset of done > perfect. โ
But it's not all about work โ I also want to make sure I'm spending quality time with my family. That's why I'm setting clear intentions and making a conscious effort to carve out time for my loved ones. ๐ก
I'm also ready to push myself out of my comfort zone and take on some new challenges. This year I'm committed to creating more content โ whether it's writing articles, being more active on social media or making youtube videos. To make this happen, I'm setting up a reward system for myself and holding myself accountable for my goals. ๐ค
So even if you're not the type to make new year's resolutions, I hope you'll join me in taking this opportunity to set some powerful intentions and make the most of the new year. What are your intentions? Take a moment to think about what you want to achieve and make a plan to make it happen. Remember, setting actionable goals and holding yourself accountable are key to making progress and achieving your dreams. Let's make this our best one yet! ๐งจ๐
Closing Thoughts
It's been an incredible year, and I have personally grown so much. From the challenges I have faced to the accomplishments I have achieved, I have learned and grown in ways that I never could have imagined. I am so grateful to everyone who has been a part of my journey, and those that have supported me. As we look ahead to the new year, let's remember all that we have learned and all that we have achieved. Let's be motivated by the knowledge that we are capable of so much, and let's be inspired to continue growing and learning. Together, we can make the coming year even better than the last, and we can achieve great things. Thank you again to everyone who has been a part of my journey, and here's to an amazing year ahead! ๐ฅ
Loved your video on GANs, excited to see what you do next year!
Love this!