For AI Democratization, We Need to Change Popular Beliefs Around AI

Published in

Obviously AI

6 min readJul 3, 2020

At this point, pretty much everyone knows what AI is, yet no one knows what AI will be in the future. In the news, AI exacerbates some societal fears and the narrative is — let’s just say bleak.

However, even though there are challenges we face as an industry, we have an incredible opportunity to overcome them. There are many popular beliefs about what AI is and how it affects you. Whatever the beliefs might be — negative or positive — the future of AI is still largely unwritten.

The way people think about AI and how it’s going to be a part of our lives has to change. And we want Downsample to contribute to the culture shift.

As of now, we have laid out some goals of this publication.

The 3 goals of Downsample:

Downsample will present ideas on how AI can be democratized, be more inclusive, and be accessible to all. We believe AI should be put in the hands of all — or democratized. Everyone should be able to use AI, yet right now they can’t because of a skills gap or lack of resources.
Amplify voices in the AI space. This means we will aim to build a diverse team of writers to discuss how to face the problems in AI. This will include closing the skills gap, making AI more inclusive, AI ethics, algorithmic bias, and more.
Change beliefs about what is possible with AI. While this is a complicated goal, we need break down what is possible with AI and how it can be better in the future.

Here Are the Popular Beliefs We Need to Change to Move Forward

So the question remains, how do we shift the culture surrounding AI?

We need to destroy popular beliefs around AI and rebuild them.

Let’s start with one of the biggest beliefs:

I Don’t Know Data Science or Have the Resources to Learn It.

This is a big reason someone might not think they can use AI themselves. What if we told you most problems can be solved if AI is democratized?

While we know it is slightly more complicated than that, we also know that if we can put AI in the hands of everyone and not just a few engineers, everyone will have the knowledge of what’s possible with AI, what makes a good data prediction, how to govern data, how to avoid bias data, and also diversify those who control the machine learning process.

If AI is democratized and the barrier to entry is lower, you can also do more with less resources. This can help startups and consumers access AI the same way large enterprises do.

Here are some ways AI is being democratized 📜

No-Code — AI tools you can use without code or a technical knowledge democratizes the technology the fastest. Just like Squarespace and WordPress did in the early 2010s in the web building space, allowing anyone to build a website, now you don’t have to be a data scientist to use AI.

Open Source — Like no-code, open source AI has a certain culture built around it. Open source is built on the belief you should give away the technology you develop to excel the software faster. You may have heard of open source machine learning frameworks like TensorFlow or companies like OpenAI aiming to benefit humanity with open source.

APIs — APIs can add machine learning or AI to an app with one line of code. It is slightly more technical, but no-code tools who build APIs for their users enable them to give their users AI, thereby democratizing AI.

You Can’t Experiment With AI or Interpret Outcomes Meaningfully.

With traditional AI being a long, complex process there’s little room for experimentation. Democratization comes into play here as well. With AI tools that use visual programming and drag-and-drop features, you can close the skills gap. Additionally, if you can cut down the time in between collecting data and getting an output, you can increase experimentation because you don’t have to worry about wasting time. For example, if you can get 20 data predictions a day compared to 1 a week, you can experiment and be creative with how you use AI.

If AI can be democratized, the AI process ideally would become more transparent, easier to communicate, and accountability would increase. Imagine if anyone on your team can explain the outputs of a machine learning model. The organization would only benefit and AI interpretability would be attainable.

AI Bias Can’t Be Avoided.

Perhaps its biggest problem, AI can be notoriously biased and unfair. A popular belief might be, AI bias can’t be avoided. If we change this belief and approach with a democratization mindset, we can start to put data governance, collection, and model building in the hands of a diverse group instead of a few technically-minded engineers.

Here’s a recent quote from OneZero’s “‘We’re Drowning in Data’: How to Think for Yourself in the Age of Experts and A.I.”

But we can’t lose track of the fact — and this is a really critical point — we can’t lose track of the fact that technologies, algorithms, artificial intelligence are all designed by humans.
Humans set the initial conditions. We may not know where those official conditions take the technology with machine learning. But we know we taught it how to learn, so to say, or we set the initial parameters, and we’re finding that in many of those cases, the algorithms are producing biased outcomes. — Vikram Mansharamani, lecturer at Harvard’s School of Engineering and Applied Sciences

If we can fix the AI space at the human level, we can start fixing the technology we build.

So how do we even begin to address such a big problem? Increasing diversity and amplifying voices in the AI space is a great start.

According to IBM:

“85 percent of AI professionals believe the industry has become more diverse over the past few years; of those, 91 percent think that shift is having a positive impact. 74 percent of AI professionals believing diversity hasn’t improved say the industry must become more diverse to reach its potential.”

AI diversity will undoubtedly benefit the industry by promoting accountability, fairness, and data governance.

You can read about some inclusive organizations in the AI space here.

And here are the Twitter handles of those leading the charge for more inclusive AI 🐦

Sorry if we missed any. 🥴

We will add more as we dive deeper into diversity topics.

The Long Future Ahead of AI Democratization

Like we said, no one knows the true future of AI. But if Downsample has anything to do with it, it can be a better democratized, inclusive, and accessible future.

We have a lot of work to do and a lot more articles to provide. Stay tuned. 🤖

📌 If you haven’t follow Downsample on Twitter.

📌 While you’re at it follow Obviously AI!

📌 If you have any ideas, submissions, insights you want to provide email jack@obviously.ai

📌 For more reading, check out this 2019 AI Index published by Stanford.

📌 Or the current controversy going on in the AI bias debate.

📌 Lastly, don’t forget to follow us on Medium!