AI in your grandma’s hands
TL;DR
Artificial Intelligence (AI) is on the rise and to facilitate the use of AI a whole lot of companies created end-to-end tools for people who don’t know how to code their own AI solutions. This trend is called democratized AI and aims to empower any user with the ability to create their own AI model.
Intro
Artificial Intelligence (AI) is on people’s tongues and has become one of the favorite topics to write about (sic!) by bloggers and news outlets. Will the robots take over and annihilate the human kind? Are we going to lose our jobs because of AI? These headlines make a splash on online news platforms. So, pretty much everyone wants to jump on that bandwagon and capitalize on the popularity of (quite often, ominously presented) AI.
Well, I’m not going to do that. I’m not going to be “another brick in the wall”. This time around, I’ll take you on a tour of democratized AI and explain what makes democratized AI commonplace.
What is democratized AI?
Just as Shakespeare wrote, “it’s all in the name” and as the word “democratized” implies — power is vested in the people and exercised by them. In this case, it’s not about voting rights but about the ability to create your own AI solutions regardless of your machine learning and coding skills. In other words, democratized AI is the spread of no-code, low code tools for developing AI models to the audience that is not technical. The target audience, are subject matter experts, domain experts such as doctors, biologists, scientists, roboticists, and regular folks who want to automate some task.
One of the best ways I came across that explains what democratized AI is, is Google’s promo video of Teachable Machine 2.0, a web-based tool for creating machine learning models.
This tool is geared towards everyone with no machine learning knowledge. Hence, Teachable Machine 2.0 may not be your go-to tool for more complex problems to solve.
What makes democratized AI so accessible to everyone?
The recent emergence of AI development tools hit the market, many of them are no-code tools. It’s thanks to these tools the expansion of democratized AI took off and is still gaining momentum.
The majority of these tools focus on computer vision, a branch of AI that uses deep learning to process visual data for high-level understanding from digital images or videos.
As a user, you create labels (a concept for a machine to learn) e.g. “dog” and “cat”. For instance, to teach a child what a dog is, you would show the dog to the child. Similarly, the machines need to be told what an object is, before it can be recognized automatically. Then, you would annotate or label (these two verbs are synonymous) images/videos.
In a simple classification task, you would assign a label (either dog or cat) to a presented image with a given pet. For an object detection task, you would draw a bounding box around these pets and assign them a proper label. You do it by selecting a label for an image or a region in the image/video through a chosen platform’s graphical user interface with annotation tools.
The labeling process will vary depending on the type of computer vision task. You can think of the labeling (or synonymously, annotating) process as drawing a picture in Paint or editing an image in Photoshop.
Assuming the user has labeled a collection of images, these platforms offer a built-in, one-click away (or oftentimes, automatic) option to start training your model without any setup or configuration. Once training starts, you can continue annotating the images to improve the model since training recedes to the background and occurs automatically. Once training finishes, you can export the trained model and integrate it into your application.
A long list of what you can find on the market of democratized AI
Many no-code platforms sprung up in recent years. I will list some of them for you to try and take a gander at. Trust me, exploring them yourself is way more fun than reading reviews without actually never using any of these tools.
Conclusion
AI is going streamline and one could say it is a good thing. Most citizens of most countries can potentially harness the power of creating a deep learning model without knowing any mathematics, coding, and data science. However, the results of model training can often be unexpected. So, an average Joe’s trained model could be the next Frankenstein but this is what we will cover in the next installment of this AI series.
