Making AI accessible

March 6, 2024

Why use JumpStory?

Save time: We've curated the best images for you.

Save money: Unlimited use for just $23 per month.

Get the latest posts delivered to your inbox


Why use JumpStory?

- Save time:
  We've curated the best   images for you.

- Save money:

  Unlimited use from $23   per month.

In this article, Louis Bouchard explores the impact that AI has on the world and discusses the need of educating people about AI.

1. What is AI (to you)?

This is a great ongoing question. To me, artificial intelligence (AI) just means the use of machines to do a task that typically requires human knowledge. Such a task includes many different approaches involving, or not, machine and deep learning, but those are definitely the currently most hyped approaches.

Expert knowledge approaches using regular coding and conditions based on knowledge from experts in the field is also a type of “artificial intelligence” as it allows a machine to answer questions or take related actions based on inputs.

2. What does it mean to make AI accessible?

Making AI accessible means two things. First, it means sharing what AI is capable of with everyone, not only professionals. AI technologies surround us, and they will only take a larger part of our life. I believe it is important to know what it is and at least understand how it works at a high level to keep the control and not be scared or have realistic expectations.

 

Man with smartphone - AI technology surrounds us

 

A second meaning is to share new research and progress in the field with anyone interested in the field in a more technical manner. This is just as important as I believe such communication helps progress in the field, allowing all researchers and professionals to be up-to-date with the different works in the field and learn from it.

3. This book is all about impact. How will AI impact the world?

Oh wow, that is a large question. I can only say that AI will, and already do, impact the world in many if not all possible ways. It is already in most industries and controlling some of our behaviors through recommendation systems and such.

You can try to think of the industry that is the least susceptible to using AI, and it will already have a use case where AI can help. Think of agriculture; well, there are already automatic water systems using computer vision and prediction systems to optimize and automate watering. Think of groceries? Just visit an Amazon Go store, and you will see how impressive a convenience store can get using AI.

4. Impact is not necessarily a good thing. How do we ensure that AI doesn’t make the world a worse place than it is now? For instance, I think that it’s still up to discussion if companies like Meta have in fact made the world better or not. Would you agree?

I believe making AI accessible is one way, if not the best, to use AI “for good”. Unfortunately, I believe it is impossible to ensure (with 100% confidence) that AI is being used to make the world a better place, but we, AI researchers, will definitely try our best to make it happen. The government is best placed to control the big companies’ impact, but they need to understand the technology itself and hire experts that deeply understand it.

Only experts will not suffice, nor will only high-level understanding either. Both are required, which is why popularizing AI is important and why sharing up-to-date information with fellow researchers also is.

 

Techn researcher

 

The government needs to predict what will happen next, and it needs the best researchers and information to do so. The person making the decision, in the end, should also have a high-level understanding at the minimum.

Lacking better examples; Would you let someone from another country with completely different education systems and zero knowledge about education at all decide if we should add a year or not to the school curriculum? Of course not. You want someone that studied the subject and understood the issues and possible consequences.

5. Most technology CAN be built, but SHOULD it be built? What do you think about this question, when you relate it to AI generating e.g., faces and people that don’t really exist?

Something that can be built shouldn’t necessarily be built. For example, we can surely build a fake Obama video saying horrible things about the United States, but nobody should do that. Similarly, we can create horrible AI-based automatic shooting drones, and some already are making such weapons, but the world would be better without them.

AI is just a tool we can use to improve our lives, but it can also be used to worsen the situation if in bad hands. This is why we need a good regulation by the government and why they need to understand that.

6. AI is getting better and better at restoring photos, but it can also generate deepfakes, where we can’t tell the truth from reality. How do we set up global guidelines to prevent the negative impacts of these technologies, or are you against such guidelines or regulations?

AI is getting better and better at restoring photos, but this is the same technology as deepfakes and doesn’t really restore a photo. In fact, if there’s a scratch or a graytone picture, the model simply learns from images in a big database what is the most expected feature to appear below the scratch or what color is supposed to be where in the case of a gray image. It will not reconstruct the original image but will rather take interpolating from the dataset.

 

Dataset

 

Here, I want to highlight that it is interpolating, not extrapolating. It only guesses from what it knows and doesn’t extrapolate as humans can do. Thus, if the image wasn’t in the database, it is literally impossible for the model to make an exact guess. Still, it will be pretty close and produce incredible results for sure!

7. Apart from being the editors of this book, we also work within the space of stock images (JumpStory). What will AI mean for this industry in 10 years from now? Will major companies like Shutterstock and Getty Images simply be history, because we don’t need them or photographers anymore?

From what we already see with image generation models like Imagen or Dalle 2, I believe many companies will create images for users using AI rather than stock pictures. Shutterstock may or may not adapt to this change, and even better models will also appear.

I couldn’t even make a guess one year from now at the speed progress is going right now, and I believe life will be so different in 10 years that it is impossible to answer this question with a legitimate guess.

 

Person taking a picture with a camera

 

8. Should all companies care about AI? What if I work as a construction worker? Do I really need to understand these new technologies in order to keep my job in 10 years’ time?

I don’t think all companies should care about AI. AI isn’t magic, and you don’t always need it, nor will it always help you or your product.

You need to get into AI and use it when you don’t find any other solutions to a problem. Still, even in this case, I believe a construction worker should care about AI but not for the same reason. If this person has a smartphone, then his/her life is already impacted by AI in many ways to ensure the algorithms aren’t controlling their behavior or at least not be scared of upcoming changes like thinking of losing his/her job in a few years.

Staying up-to-date and understanding the current state of AI will maximize your well-being and minimize stress regarding this technology. In fact, AI-powered robotic arms and tools might replace construction workers a few years from now, but we don’t know when. It is definitely possible and not that far away.

9. Machines are not very good at ‘out of box thinking’. Is this also something that they will become better at in the future? Meaning that they can do things that they are not designed or programmed to do?

As I said previously, machines are currently merely interpolating from data and not extrapolating from it. Meaning that it cannot “create” new things but only merge and blend existing things. Even the hottest models like Dalle that generate amazing-looking images will not create something new. Instead, combine existing concepts and images together to make it look unique and cool, but not creative nor new.

It is definitely possible that AI will be designed to extrapolate and even be conscious one day, but we will have to find a completely different approach to achieve that and not only keep scaling to larger datasets and faster inferences.

About the author

Louis Bouchard is a researcher, YouTuber, and community manager. He is co-founder of Toward’s AI and owner of “What’s AI”.

Louis Bouchard

How useful was this post?

Click on a star to rate it!

Average rating 5 / 5. Vote count: 23

No votes so far! Be the first to rate this post.

Want more?

Get 10 free, down-to-earth eBooks about marketing, visual communication and storytelling. Written by authors like Guy Kawasaki (Apple), Lars Fjeldsoe (Dropbox), Peter Kreiner (Noma) and more.

Want 25% off?

You're likely not ready to try our service right now. And that's alright.

How about we e-mail you a discount code for when you are?

We will of course also send you a brief explanation on what’s so special about us. And subscribe you to our regular newsletter, where you’ll be notified when we publish content like this page.

But don’t worry. We won’t share your address with anyone else. And it’s just one click, then you’re off our list again.