The new wave of AI-generated art and design being produced by Dalle-2 is undeniably incredible, and where it truly excels is in the area of ideation, creating an array of realistic images from simple text descriptions in a matter of seconds.
Frankly, compared with the traditional ways that artists, designers, or engineers undertake these sorts of tasks, it’s revolutionary. Not just in its speed of execution, but also the quality and variety of ideas that are produced.
Is it perfect? No. But it doesn’t need to be for it to fundamentally impact the way the creative process is undertaken. I think anyone that encounters AI-generated art recognises this, which is why there are thousands of hot-takes about Dalle-2 being simultaneously devastating and exciting for creative professionals.
Honestly though, I think attempts to paint Dalle-2 as a threat are just rehashing tired tropes about the existential threat of technology. As a designer I take an unashamedly optimistic view on this new technology, and rather than joining the debate I thought it would be nice to showcase just a handful of the most inspiring examples I’ve seen recently.
AI ideation in action
Film & television cameos
Wondering what your favourite character might look like with a guest spot in another movie or show? Maybe you’re pitching to studio execs about a crossover episode, or just making a fun print for your friends birthday: it doesn’t matter really, Dalle-2’s got your back. Scroll through this great thread from @HvnsLstAngel on twitter and you’ll see what I’m talking about:
Would Mondriaan-inspired Sneakers work? I’m not sure, I can sort of imagine it, but I really don’t want to spend my time browsing reference pictures and drafting sketches to find out. Enter Dalle-2, courtesy of @GK3 on twitter:
What I love about this is that Dalle-2 has seemingly had to accomodate the constraint that shoes just aren’t made up of the squares and rectangles we’d typically associate with Mondriaan. Instead, it’s chosen to adopt the classic Mondriaan color palette and embellish the designs with artboards containing rectangular Mondriaan-inspired backdrops. It’s stunning, as are many of the other mashups in George’s thread.
Bonus addition: Here’s another beautiful sneaker experiment I spotted just before publishing:
Thanks to the OpenAI team introducing us to the avocado chair, the fruity furniture theme has become a familiar idea within the AI-generated design space. That said, it doesn’t make it any less impressive. You can find countless great examples on twitter, but I particularly liked some of the outcomes in this thread from @RAC
One application of Dalle-2 that you see time and time again is digital art. The unique thing about using fine art as a creative parameter is that it’s so inherently open to interpretation. We’re more inclined to accept the visuals as human-generated because we’re already accepting of the variations in artistic expression.
There were hundreds of threads I could have chosen to share here, but I really loved how @shashj’s experiments use Dalle-2 to combine traditional art styles, such as Mughal painting or Japanese woodblock prints, with modern concepts like quantum computing and nuclear fusion.
The greatest strengths of Dalle-2 at present often lay in expressive design, ambiguity, and fantasy - stuff we can imagine being real, but we need to apply our own creative interpretation to help them succeed in our mind’s eye. For example, seeing AI attempt to render typography can often shatter the illusion of intelligence, as it struggles with typesetting and generating believable words. That said, where past models failed, Dalle-2 is begining to succeed, as these examples from @jmhessle illustrate:
There’s a whole host of AI companies that have sprung up over the last 5 years that help fashion retailers to digitally replace outfits on stock models or even computer generated super models. Well, Dalle-2 can go a step further, and invent and/or swap out clothing items on the fly:
New styles of (Ai)rt
Something about these examples above, no matter how stunning, remains derivative. Art is always derivative though, and the compositions are unique, even if you can intuitively place them in existing bodies or work or aesthetic traditions.
But what do we call some of this art that looks entirely new, art that we can’t really place in one tradition, art from machines but augmented by our own creative hand? Perhaps it’s (Ai)rt or Human-AI-collaborative practice? I don’t really know what we’re going to end up calling it, but I’m starting to know it when I see it.
In the examples below you can hardly deny that things like ‘Dali’, or ‘Hieronymus Bosch’ were used as input parameters:
But the composite that @_dschnurr has created in the 3rd image (below) feels different. I mean sure, it’s surrealism and collage; it’s kind of Dali meets Jim Warren meets NoMan’s Sky meets 80’s video game aesthetics, but it’s also arguably something new. I’m not saying someone couldn’t have made this image 25 years ago on Photoshop, but the fact that it’s a true human-ai collaboration is symbolic of the new era of creative practice we’re entering.
If design is my livelihood, it’s perhaps worth concluding with an explanation as to why am I so optimistic.
As we’ve seen time and time again, when a new technology arrives that threatens to upend the established ways of working it’s common to see people ringing alarm bells - fearing mass unemployment and the loss of traditional skills, or pointing to a dark new future where we’re completely replaced by machines.
As human technological progress accelerated during the industrial revolution the world changed at a speed that nobody had ever encountered before, and it was reasonable that we might be fearful of what lay ahead. However, since then, the rate of innovation hasn’t slowed down. We’re now two and half centuries into an industrial and technological revolution that shows no sign of slowing down, and most importantly the human experience is markedly improved.
We now know, as a matter of historical fact, that new technology primarily leads to more innovation, and net-postive outcomes for workers and societies. Sure, the way we generate and execute on new ideas is changing, but it always has, and there are still craftspeople and designers and engineers.
If anything, our creative lives have become more rewarding as we augment our creative practices and expression using whichever technologies are on hand. Maybe I’ll change my tune when AI can design better software than me and I’m out of a job, but as a designer, how can I not be excited about seeing what comes next?
Update: Things have changed rapidly since we published this article, so we've written a follow-on post explaining why we think the growth of text-to-image creative tools is so explosive at the moment. Click here to read our next article on AI-human collaboration.