In 2016 the Internet went crazy about speaking machines. What we couldn’t predict was that 5 years later, in 2022, similar machines would be creative too, or at least able to take human speech and transform it into images as fascinating as these.

Trying DALL·E

From brief to versions in 10 seconds

Both images are two of the first four suggestions I got back when I asked DALL·E to create a “dark blue and soft pink abstract geometric image that represents progress, technology and humans.”

Of course, there’s nothing super innovative in what DALL·E did. I could have asked a graphic designer to create two images for me. They could have worked with their own models, or they could have used a stock image.

What’s interesting is that DALL·E did in 10 seconds what a graphic designer would have taken hours, if not days or weeks to produce.

DALL·E gave me 4 options in 10 seconds: Two with humans, and two abstract images, with different styles and color palettes.

And then, when I asked, 3 more variations of the ones I liked the most, again in seconds.

Skip the hard part, keep the rights

With DALL·E there’s no model, there’s no photographer, there’s no waiting. None of the humans exist, and if you remove the need to take pictures of real people you can save a lot of money.

I generated these images in 10 seconds. And nobody had to pose for them. These models have no image rights, charge no modeling fees and drink no coffee. They don’t need outfits, or lighting, or toilet breaks.

With DALL·E, scouting is more about being great at defining the characteristics you want your models to have, and less about being able to afford a great agency or model.

If you’re wondering about originality, none of the images have ever been generated before or will ever be generated again, even if you ask DALL·E to generate something for you with the same request.

From oil paintings to portraits to UIs

Another thing that’s interesting about DALL·E is that it’s able to create different styles and techniques. With a typical designer or creative, the process begins with looking for someone that has a style you like for your project. But DALL·E doesn’t have a style. DALL·E can make realistic compositions, paint abstract oil paintings, logos, user interfaces, and anything you can imagine.

I asked DALL·E to paint an “abstract yellow violet oil painting”. I’m no artist or expert, but these two options look good enough to me.

I asked DALL·E to paint a paint monster drinking coffee and an old cat lady drinking mate in a living room. Check.

And look at these mockups. I don’t know about you, but I’ve always admired creatives who can do more than one style. And I was wowed by DALL·E.

Bits of the future

As Genevieve Bell said in this great talk, bits of the future are already around us. DALL·E and the rest of AIs are a great example.

DALL·E is not alone. Other Artificial Intelligence programs can paint and design; some like GPT-3 specialize in writing, some like Github’s Copilot and AWS’s CodeWhisperer help developers write better code, and hundreds of specialized AIs already help doctors spot issues in medical imagery that may escape the human eye.

In the future, a lot of what we do will be done in cooperation with machines. We’ll have to learn how to ask for things and how to refine our queries to get the results we want. We’ll have to develop our taste to select the best options from all the suggestions we get from our AIs. Lawyers and intellectual property experts will need to catchup to deal with new rules, and new rules will need to be written.

I believe there’s a while to go for these AIs to take our jobs (or many o them at least), and that there will always be room for specialists and something unique and special about what’s created by humans.

I don’t think the debate is around whether AI will replace a lot of what we do. I’m sure it will.

AIs can save a lot of time and money, and savings is a magic word that will motivate a lot of organizations to move some of their workload to AIs. While non-digitalized fruit shops and design shops with an artisan mentality will always have a place in the market, assume that AIs will become ubiquitous in many businesses, and as common as spreadsheets and chat software.

I believe the question is how it will happen and how we are getting ready for it. What can we do for AIs to join the workforce (and our societies in general) in the best possible way for humankind?

Why the human in the loop?

If you’re wondering what’s for us to do with how AIs influence our world and why it’s important for humanity, here’s an example.

The know-it-all artist able to create beautiful oil paintings and good-looking user interfaces in seconds believes this is a realistic boss. Based on the training data, DALL·E believes that bosses are men who are willing to wear a tie. Unless a human correct this bias, this problem will continue to replicate, and will cause the materials we create using DALL·E’s images to be biased, which will make more training data biased… you get the roll. And only you, or another human, can stop it.

The bright side

DALL·E has been programmed to block the generation of images with violent or sexual terms, which is a great step, and continues making changes to ensure results are safe and diverse, but there are biases in DALL·E (just like there are biases in Google Translate and many other tools we use today, on the other hand).

OpenAI, the organization behind DALL·E, has been open about the biases and publicly recognized that bias from their training sets affects their results.

Spotting and correcting the bias is a good reason to ensure the future of our relationship with AI is not of competition, but of collaboration. Machines can augment our imagination, but they need our human touch to get to the best results and make sure results are inclusive and don’t drag other efforts on diversity. Get ready to become great at collaborating with machines, the future is already here!