Is the democratization of AI good?


This post was originally published by Ayaan Haque at Towards Data Science - Medium Tagged

An opinion and discussion piece on whether AI democratization is on net beneficial

Image from Unsplash by Markus Winkler

In the modern age of education, almost anyone with an internet connection can learn anything they want to. This is also true for learning AI, and now, anyone with the requisite background has the opportunity to learn AI and build AI programs. When I say “democratization,” I mean the easy access to AI education and learning, and more importantly, the easy access to building scalable AI applications. In an article I wrote earlier this summer, I discussed my personal experience with AI ethics and how I paid little regard to the implications of my work.

That article is here:

I have always heard that the democratization of any sort of learning is beneficial, which I generally agree with. This has also been extended to AI, and for the major part of my AI career, I have completely agreed. I am a high schooler who didn’t start a calculus course until just 11 months ago. However, up to that point, I had made a lot of progress in AI simply because I could learn from free Youtube videos and projects online. However, at this point, after considering how I and others have used AI, my strong opinion is somewhat wavering. While I honestly don’t have a solidified opinion on the topic, in this article, I will discuss what I think are some of the main benefits of AI democratization, and more importantly, what some of the downsides are since I believe many people don’t talk about it.

I’ll keep this section short. Obviously, there are so many reasons why having AI so accessible to so many people is beneficial. First, it provides people opportunities to learn and build careers that they never had before. For people who may not have access to formal education, bootcamps, Coursera courses, and Google’s certifications are great alternatives for people to get into AI and find job opportunities. Moreover, it allows younger developers like me to build an interest in AI and pursue it in college.

Most importantly, the democratization of AI creates a large wave of AI engineers, researchers, and developers for the next few decades. This is important, as only new in history are we able to teach such complex concepts to such young people. This should undoubtedly accelerate the growth of AI and help it be integrated into more parts of our society. However, there are concerns over how fast we integrate these new AI systems into our daily lives. While I can name many more benefits, I will stop here because I think you get the idea. What is more important and less discussed in my opinion are some negatives of the easy access to AI.

There honestly are not many downsides to the accessibility of AI, however, the major and obvious one would be ethical concerns.

Being a young developer, my work is often seen as gimmicky in the AI community. While veterans in the community often overly praise me for my work, sometimes they don’t take my understanding and work seriously. This leads to major problems when I try to launch real apps or make a real impact on my work. As I described in the article I linked at the top, many people reached out and warned me about the ethics of my work. The problem was that I didn’t understand the ethics and didn’t care to think about it.

One major issue is that almost every beginner AI course, including my own, discusses areas where AI can be implemented that have major ethical issues, such as healthcare. For younger kids who are beginning to learn AI, they believe that they can create AI systems to use in real systems, making them get carried away. For eager and driven young minds, they won’t take no for an answer and will go straight into a project without considering the harms. This is exactly what I did. This is why I fear democratization, as it puts AI in the hands of people who may not understand what they are building and why their work is potentially dangerous. While this is only one of the few reservations I have, it is a very major one and should be considered.

So this prompts the question, when should people start being exposed to AI? I believe it can be early in college, but honestly, this is a challenging question to answer. A much better solution would be to include multiple ethics lessons in any AI course or curriculum. It would have to be one of the first lessons in a course so that students learn exactly what AI ethics are and why they are so important. This would enforce caution and responsibility for students, which would solve this problem.

The other main concern is regarding a lack of understanding. A quick look at Coursera and you can find many AI courses that focus only on development and no theory. This means that newcomers learning AI are only learning how to code neural networks in Keras and PyTorch without understanding any of the theory that goes behind it. I fell victim to this trap very quickly, and I hope to write about that experience in the near future. This is a major problem for the AI community because it creates engineers without a proper technical background. If the theory isn’t taught properly in the beginning, it is much harder to enforce it properly later, as it is a foundational block. Therefore, once new AI developers reach the job market, they are weaker candidates and if they are hired, the progress of AI is halted. We need to teach new developers properly and efficiently in order for AI to progress at the rate it needs to.

These are the two major negatives for the democratization of AI, mostly because I have seen and experienced these first-hand. If you have any other concerns, let me know in the comments so we can talk about them.

I honestly believe that the democratization of AI is on net beneficial. I wouldn’t be able to write this article if it didn’t exist. I am literally only 17, and just a few decades ago it would be unfathomable for such a young person to be knowledgeable about such an advanced topic. But now, in 2021, I am hardly unique and there are students younger than me who know even more than I do. I think the major problems that we face regarding the democratization of AI can be addressed somewhat easily, it will just take proper realization and effort from those who have constructed this institution. But for now, I will do my best to prevent myself and those around me from falling for the traps that will halt the progress of AI. Hopefully, this article can reach the right people and help create a mindset change.

Spread the word

This post was originally published by Ayaan Haque at Towards Data Science - Medium Tagged

Related posts