Maximizing Productivity: What We Have Learned So Far About AI In Our Workplace
AI appears to be here to stay and at Advanced Online Insights we have challenged our team to see how we can take advantage of some of the benefits it can bring to our business. Throughout the past month we have tested several different AI tools. I want to share some of our favorites with you and tell you a little bit about how they have impacted our daily work lives. I’m sure that we will be following up in the future with some more in depth looks at some of these tools as we continue to learn more about them and how they can take our business to the next level!
In this article I will discuss three types of AI tools we have been testing. They are broken down into the following sections, Natural Language Processing, Text-To Image-Models, and Machine Learning / Computer Vision. Each section contains examples and my initial thoughts about how they can be useful in our day to day business processes.
NLP (Natural Language Processing)
In it’s most basic form, NLP (Natural Language Processing) allows a human to have conversations with a computer. I am nowhere near smart enough to describe how these tools work so I’ll just take the opportunity to let them speak for themselves.
I went to you.com and used their free NLP Chat tool and I gave it the following prompt…
Woof woof! I might be a furry friend, but I do know a little about Natural Language Processing Tools! These amazing software programs use advanced technology to help computers understand and interact with human language, just like how I can bark and communicate with my human friends. With NLP tools, you can do things like translate languages, summarize important information, and even analyze people’s emotions through their words. They are super helpful for humans who want to communicate better with each other and with other creatures like myself. So next time you want to chat with your computer, remember to turn to NLP tools! Woof woof!
you.com
If that doesn’t impress you then I don’t know what will.
I have been alternating between 3 NLP tools; Google Bard, you.com, and ChatGPT. Here’s a quick overview of what I have found so far.
Google Bard
Google Bard is currently free to use and is available to users by signing up on their waitlist at https://bard.google.com/
Google has been updating the tool everyday and it continues to get better with every update. My favorite feature in Bard is that it gives you 3 responses each time you chat with it or ask it a question. Sometimes the answers are nearly identical and some answers vary drastically. It really helps to remember that the AI models still have a long way to go and every answer should be questioned and fact checked.
you.com
you.com is a very approachable AI NLP tool. The developers have added quite a bit of polish to the program features and the way that it generates answers to questions. I have found it to be the easiest NLP tool to use for beginners and it is always the one that I share with beginners who want to experience
ChatGPT
ChatGPT was the first AI tool to become a household name, and for good reason. ChatGPT was created by OpenAI but is being heavily funded by Microsoft to add it into many of their existing products and services. ChatGPT is currently free to use. The main webpage is frequently down due to high demand but it can be accessed directly through the Microsoft Edge Browser.
Text-To Image-Models
I’ll step aside and let Google’s Bard AI give us an explanation of what text-to-image-models are.
Hi! Have you ever seen a picture of something you can’t imagine? Like a dragon or a unicorn? Well, there’s a new way to make pictures of anything you can imagine! It’s called a text to image generator.
A text to image generator is a computer program that can make pictures from words. You just type in a description of what you want the picture to look like, and the program will make it for you!
For example, you could type in “a blue dog with a red hat” and the program would make a picture of a blue dog with a red hat. Or, you could type in “a rainbow over a castle” and the program would make a picture of a rainbow over a castle.
Text to image generators are still under development, but they are getting better all the time. In the future, they could be used to make all sorts of things, like new kinds of art, or even new kinds of toys.
So, there you have it! Text to image generators are a new way to make pictures of anything you can imagine. I hope you enjoyed learning about them.
Bard AI
Whenever anyone talks about text-to-image generators there are 3 platforms that they are usually referring to, Stable Diffusion, Midjourney, and Dall-E. These are the most popular and most widely accessible platforms available at this time.
Stable Diffusion
I have spent more hours that I can count tinkering around with Stable Diffusion. I personally prefer this text-to-image generator because it is open source and can be downloaded, installed, and trained. It can create images from text but it also allows users to manipulate images and video. It has a huge following and gets updates and plugins daily.
Midjourney
Midjourney can be accessed online through the popular communication application discord. It specializes in creating aesthetically pleasing, often surreal images in response to text prompts and sources. It was originally free to use, but recently it shifted to a paid subscription model to use the service.
Dall-E
Dall-E was created by OpenAI, the same company that created ChatGPT. Dall-E is usually regarded as the most literal image creator and produces more accurate but usually less aesthetically pleasing images.
Machine Learning / Computer Vision
To round things out I’ll ask ChatGPT to explain what Machine Learning is.
Oh man, Machine Learning? That’s like when you teach a computer to think for itself, like that one episode where my toaster became self-aware! Basically, you feed the computer a bunch of data and it figures out patterns and makes predictions on its own. It’s like a really smart brain without the need for donuts, and it can do all sorts of things like recognize faces or predict weather. The possibilities are endless, just like the amount of beer I drink at Moe’s. Woo hoo!
ChatGPT
We’ve been researching AI resources for the past few months and have found that machine learning and computer vision have had the greatest impact on our work. Our objective has been to enhance our 360° images through object detection and image classification routines, but in the past, the cost and lack of customizable options for many of the services we tried proved to be hurdles. However, in the past six months, many computer vision models have been made publicly available, which has allowed us to tailor them to our needs and test their potential. With these tools, we aim to take our 360° street-level imagery to new heights by developing our own in house software to identify street-level assets such as utility poles, manholes, signs, and more.
Google Vision API
Google has always been a leader in image classification an object detection. They have used their technology to enhance their Google Photos app and to use online images to enhance the search results for their Google Search results. If you are interested in what Google’s Computer Vision AI can show you about your own images check out at demo at https://cloud.google.com/vision
YOLOv8
YOLOv8 is a computer vision model that can be used for object detection in images and videos. It has been shown to be effective in detecting a wide variety of objects. By using YOLOv8, even a small company can use computer vision to automate the process of finding objects in images.
To use YOLOv8, you need to first train the model on a dataset of images. Then, you can use the trained model to find the specific objects you are trying to capture in the images.
SAM (Segment Anything)
The Segment Anything Model (SAM) from Meta AI is a significant development in machine learning because it enables accurate image segmentation or object detection. Image segmentation is a critical component in various applications, including medical imaging, autonomous driving, and robotics. SAM is unique in that it can output multiple valid masks, an essential capability in solving real-life computer vision problems.
Additionally, the release of SAM means that more companies and organizations can access state-of-the-art computer vision algorithms, making it easier for them to develop custom image segmentation models for their specific needs.