Adobe MAX and Text to Vector Generative AI

FRIENDS

Reuniting with fellow creative friends at Adobe MAX and forging new connections is always the highlight of any event like this. Hanging out and having meaningful one-on-one conversations with the peers you respect but don't get to see every day are the moments I enjoy most.

THE NEW FRONTIER

In a realm where diverse workflows intersect, the focal point of this year's MAX was the integration of artificial intelligence.

With AI integration into the creative landscape of Adobe Illustrator, we're sure to see a lot more crappy vector art moving forward. The 'Text to Vector' generative AI feature should be titled 'Text to Clip-Art.'

And all the reels you'll see on social platforms for a good while hyping it won't scrutinize the unoriginal results of sloppy vector art it produces.

TEACHING AND LEARNING

For educators and mentors who teach about design and illustration using Adobe Illustrator, it's imperative to guide students away from a reliance on generative AI in their assignments.

It's a disservice if you let them use a prompt-based methodology. They need to learn and nurture their foundational creative skills. Please encourage your students to develop their vector-building methods by learning to create their artwork from scratch. That struggle is how they will improve their core abilities and craftsmanship while producing a well-refined creative process based on original ideas.

AIRONY

The irony of the 'Text to Vector' generative Al in Ai (I think the name of Illustrator needs a new abbreviation to avoid confusion) is that the imagery cobbled together by algorithms is based on a database of pre-existing artwork pulled from Adobe Stock. This source imagery took skill and craftsmanship to create it initially. But the derivative result (mashup) is less refined and subpar, lacking any redeemable traits.

In a digital age, analog skills like drawing are more critical now than ever in our industry. But so many avoid it. I think the audience who will be drawn (pun intended) to 'Text to Vector' generative AI will be non-drawers. But the irony is they are creating derivative work from those who have learned how to draw to facilitate the AI algorithms.

TIME IS MONEY

Time is valuable. Can AI generate numerous design concepts quickly? Yes. Are they high-quality with impeccably precise vector artwork? No. Some may argue, "But think of the time saved during the brainstorming phase." However, cleaning up AI-generated artwork and design will require significant time that could have been better spent creating your idea from scratch.

COMPARISON

Many people at MAX approached me and asked me what I thought of the 'Text to Vector' generative AI. All I could say was that it made me feel unsettled. I had to give myself a few days to think through it, try it myself and write this article.

I knew I also needed to do some comparisons as well. The samples shown below are several comparisons to actual designs I created and the prompts I used in my attempt to see how AI stacks up in comparison to human ingenuity.

COMPARISON CONCLUSION

From my own experience generative AI in Photoshop to fill in a background, extend an image etc. seems to work very well and produces elegant results. But when it comes to anchor points and paths rather than pixels generative AI in Illustrator feels like clip-art files from the mid 1990's.

UNRESOLVED BUGS

In 2012, Adobe released Illustrator CS6 and introduced a snapping bug. I've been requesting a fix ever since.

2019, I was invited to Adobe Illustrator headquarters in New Delhi, India, for an Adobe Summit. It was one of the best travel experiences I ever had, and Adobe was super hospitable to all 14 creatives they invited.

While in India, each of us was assigned a group of engineers, and I showed them several bugs in person, and they experienced each bug for themselves firsthand. One engineer was able to fix a bug regarding expanding scatter brushes overnight, which was awesome.

The other was the snapping bug, which will now be 12 years old soon. Snapping in the context of vector art would be equivalent to what braking is on an automobile. It's a fundamental functionality for vector building that is still broken. I hoped this bug would be fixed in the latest version shown at Adobe MAX, but yet again, it's been ignored.

Instead we get AI-generated clip-art. ??