Can AI help design our mascot? | Hacker News Reader

Can AI help design our mascot?

labelf.ai

31 pointsyboris41 comments

The answer is, no. That AI cannot help design your logo.

It's putting out sophisticated, detailed images that portray an object in three dimensions, in abstract brushed styles.

You absolutely do not want such a thing as a logo. Maybe it can help with brainstorming.

Logos are mostly about geometric simplicity. A good logo is defined by a few Bezier curves and other 2D vector primitives, and uses only a few colors.

Also, in spite of the proliferation of color printing, displaying and copying, in this year 2010, a logo still has to work in black-and-white. That is for reasons not only having to do with the persistence of black and white printing and display technologies. For instance, if a logo works in black-and-white, you can render it in a brushed aluminum case, using that material alone.

Brainstorming is a great function for AI that seems overlooked. Most of what it generates is unusable directly, but can present variations that wouldn't be imagined otherwise.

Exactly! Curation is underrated.

You are very much making a point about the current design zeitgeist, not evergreen principles. More complex, 3D logos were common as recently as the 2000s.

To clarify, I'm not saying that 3D is off the table in a logo. Obviously there are 3D logos which are effective.

I'm making a general point about logos from around the past 100 years or so.

Generally, the only logos that last many decades are simpler ones that work well in black and white, and with an accent color. e.g. Nike, GE, Apple, Disney, McDonalds, YouTube, Volkswagen, etc.

Mildy stylized text of the company name is the other that lasts. e.g. Coca-Cola, Nintendo, Sony, Honda, Home Depot, Kellogs, etc.

I can't think of a company off the top of my head that is currently using a complex 3D logo right now.

Quaker Oats is one. (Not disputing your point; was just trying to think of any memorable 3D logos, and it came to mind.) They tried to go with a simpler, monochromatic logo and ended up switching back to the 3D drawing.

You are right, but very traditional. Try to think in a modern spirit.

If the logo can be created for by a few hundred bugs of ai service there is no need for it to last longer then a few months. Also traditional logos were created for print. If you are a 'onscreen only' company, there is really no need for a simple logo. (dont bother about asset size when the average website includes several mb of spam and js frameworks)

Further I would not know why a ai would eventually not be capable of generating a 'simple' logo with just blezer curves and few colors.

Lastly I think many of these 'simple logos' are boring. I admit I prefer the ai logos from the post. Yes call me a barbarian and yes the human drawn logo's could probably be better if made by a top notch designer (and costs $$$$$).

The future will probably be a hybrid situation where the designer is aided by ai.

"then there is no need for it to last more than a couple months" - this isn't true at all, a long lasting logo is something that is very important to branding. If you keep changing your logo, customers won't form opinions on your brand it will simply be left in the memory hole

Compassion — not compression — is a big part of design and creativity; coming up with original and unexpected results. What we see in these explorations is that AI is, I think, good at generating a bunch of novel yet visually repetitive and nonsensical outputs. The helpful-looking-elf-santa-hat-robot iterations from AI all look like variations of 1 concept with their monotonous use of gradients and brush stroke styling. Good for brainstorming? A good human designer, based on these same instructions, would ideate and generate a wider range of concepts. And we see that in the results from humans, which are actually much more interesting and diverse in terms of contrasting styles, their use of form, colour and negative space in my opinion.

Curious, was the inspiration for the Labelf AI logo the Wall-E animated movie character Eva?

Disclaimer: not a designer. Personal opinion below.

While these AI derived "logos" may have a place in design flow, they are far from a suitable replacement for human design. The images lack the fundamental creative symbolism that a human designer puts into their designs- there is a very careful method that graphic designers follow to design icons and advertising. Small graphical details are typically symbolic of the company/product line and message. An AI is going to struggle hard to deliver this.

I do see AI design as a useful tool in brainstorming and roughing designs - having infinite variations on a theme to play with can be very useful, just not as a standalone process.

This looks like an AI version of "design by committee", where a bunch of wanted features are merged into a hyena-like monstrosity.

Reminds me of "Twipsy", the mascot of Expo 2000 in Germany https://www.expo2000.de/index.php/expo-2000/twipsy.html

I am really struggling with whether or not this is satire.

Some of the AI generated ones just need a little bit of a tweak and I think they’d look a lot better than the human generated ones. It seems like the AI ones would be a good jumping off point to give humans to go off of rather than a badly drawn sketch as a starting point.

Based on what I saw, looks like the answer to the question is "no".

Yeah, none of these logos, human- or AI-made look usable (unless you don't care about logos, in which case why even bother).

The AI ones look like some sort of inbred robots

Their current logo looks like Eve from Wall-E in a pointy hat. "Good artists create, great artist copy", but that's a bit too verbatim.

The "future of infinite content" is unlikely to emerge. Procedural generation has been around for a long time and it's almost useless without curation. Creativity isn't an objective function (in either sense of the word), because it's necessarily subjective (function depends on the observer) and dynamic (function changes over time).

Finally, a blog post that answers the question 'what if Francis Bacon had been a commercial artist?'

It's a fine article with an unusually deep dive, but I'm particularly fascinated by how the human-drawn logos have a clear, minimalistic visual grammar while the AI offerings are strangely impressionist - many unsuitable, but some excitingly different.

Francis Bacon was actually part of some prompts haha

You have endeared me to your brand of techno-dystopia.

I’ve never seen a Fiverr produced logo that was non-horrible. The AI ones are equally bad.

If you have no budget then set your name in a nice font, pick a color and call it a day.

If you have a $50 budget then you can use a commercial font.

You end up looking just fine.

If your startup gets some traction and you want something more, then find a freelance designer on dribbble and hire then for $1000+

These are conceptually interesting but like others said they don’t really fit the purpose. They’re almost all unreadable at smaller sizes.

My personal view on it is (likely) as far from making logos as it is from telling stories. That is at least one but maybe multiple paradigms away.

Yeah, it can help with design but it cannot do it end to end. I do however think it is possible to create a tool that could make a good logo with more input from the user. The output of biggan is not properly suited to produce a vector logo as an end result, therefore it can only be helpful for producing concepts/ideas.

Im quite confident that something like dall-e and various next-gen clip models together with more controlled user inputs such as make this eye smaller or I want an eye here, a mouth here and add a proper way to render vectors with color choices it would be possible to get a result that would need very little polishing. We did this back ion january/feb when clip and dall-e had only been out for a few weeks and that was the first real big sign that something like this would be possible.

Heres some more recent experiments i did last night with letting a model complete a design.

https://twitter.com/ViktorAlm/status/1428094471200202758

The answer is yes. Maybe not design the whole logo, but it can definitely help as this post has shown I personally think the AI generated figures are more visually appealing compared to the human made ones, even though they may need a bit oof tweaking around the edges

It seems like it’s a great way to get ideas. A lot of art (for some artists at least) is a little bit of throwing stuff on the wall before your taste spots something forming. That’s when the ‘human’ artist kicks in, but everything before that is a lot of just trying shit.

Did the author pay to use those human-created examples? I have a feeling that if they didn't, the freelancer site probably has some fine print against using them.

given what AI does today, why in the world would you want a regurgitated mashup of all previous logos?

How is that original or memorable?

But most commercial artwork is exactly that. Not the all time great stuff, but that’s a small percentage of the market. If you need “illustration of person typing on keyboard for my SaaS website” there’s no reason that software couldn’t create that.

I forget who said it, but I once read some great design advice which was 'be ahead of your time, but not too far or people won't get it. 20 minutes is about right.'

Freaky. They look like drawings of things, but upon closer inspection, they're nonsensical.

What if I want concept art for my card game? Would this AI be helpful in that case?

Yes.

Mascots are not logos!

I'd bet if they changed the blog post to "Can AI help design our mascot" - all the comments on HN would switch from no to yes? (I'm not a designer tho)

awful AI renders.

well yes, its somewhere way out in BIGGANs space, its expected to be.

Nah, the AI ones are better, but look at the variety in the human designs, how the components are direct in their abstraction instead of, well a blob where none of it has real meaning.

ok now compare these to hiring a professional designer

Reply on news.ycombinator.com