3 ms·
> So all of that to say that it’s very, like very, difficult to describe an image with words. DALL-E isn't going to steal your job... but likely within a year
by crazygringo 4y ago
> So all of that to say that it’s very, like very, difficult to describe an image with words.
DALL-E isn't going to steal your job... but likely within a year a better interface for it will.
I don't think it's going to take too long to be able to specify textual components for image parts to occupy certain areas, and then to do iterative non-textual refinement simply by selecting among rendered variations "genetically" until you get what you're looking for.
> It took me maybe 10 to 15mins to generate all those images... And it took me something like 30mins to draw it myself,
But the point is, I can't draw. We are very quickly going to reach a point where anybody can produce basically the exact illustration they want in 15 min. It's not here today, but will almost certainly be here in 2 years if programmers and designers are allowed to experiment with the raw models.
I actually think discovering the best UI interface for generating illustrations, the magic combo of prompts+seeds+outlines+refinement, is going to be the main area of progress soon.
- adriand 4y ago> But the point is, I can't draw. We are very quickly going to reach a point where anybody can produce basically the exact illustration they want in 15 min. I’m curious about the economic impact to illustrators, artists and designers because while I agree with you that this opens up the ability for people who can’t draw to produce stuff they couldn’t before, it seems to me that this is a fairly marginal use-case. I’m personally excited about these tools because, like many electronic musicians, it’s helpful to be able to produce cool album art easily. But I wouldn’t have paid good money for this before. At best I’d pay some freelancer on the internet $20 to throw something together. Meanwhile, in the professional contexts where I’ve dealt with profesional artists and designers, the demands are far too precise and nuanced for these tools to be really useful. I understand that these tools are still in their early stages, but is it possible that, like self-driving cars, the last mile in terms of professional-level suitability will be much more difficult than we might expect?
- mlboss 4y agoWe cannot compare self-driving with art generation. If self driving fails at an edge cases somebody is hurt or dies. When art fails you just need a better prompt.
- snowwrestler 4y agoWhere self-driving and art generation are similar is a belief in the extremes of computing capabilities. Self-driving enthusiasts believe that a computer will be able to operate a car far more safely than a human. This is in spite of the complete lack of evidence (so far) that this is true. Similarly, look at your statement “when art fails you just need a better prompt.” It seems to exclude the possibility that computers will be incapable of creating certain art no matter what the prompt is. I’m not saying this because of some ineffable or supernatural attribute of human creativity, but because of the shortcomings in our flawed human attempts to create computing systems.
- b3morales 4y agoBut are those shortcomings fundamental and insurmountable? I don't believe so. While I agree with you about the current state of the art, in principle I think the project of "AI creates original art" is possible. Though it may take a few centuries to get there.
- hgsgm 4y ago
- dougabug 4y ago> But the point is, I can't draw. We are very quickly going to reach a point where anybody can produce basically the exact illustration they want in 15 min. Exact is an overly strong word. Suppose you could put onto the page exactly what you thought a pterodactyl looked like. Would you be surprised that there were skilled technicians who could produce much more accurate and visually appealing illustrations? You may not even know what you actually want, until after you see it rendered by a skilled hand.
- sinuhe69 4y agoYes, like other technologies, DALL-E and Co. will change and empower the mass, like the compact camera and the phone camera before. It feels like a tectonic change because of its potential impact for the mass. But most people who can not draw have also never hired a professional to do some work. It might make the old, not demanding works obsolete. But not the top 5%.