Lets talk about DALLE-2 A.I. text to image generation, and if we should be worried or not

General Forums Blender and CG Discussions
1

DALLE-2 You probably all know about already is a new A.I. text to image generation system build by OpenAI… And rather ironically it’s not very open source at the moment its only available to a select few screened beta testers. It creates images of astonishing artistic quality, creativity and originality in about 10 seconds based upon a text prompt. It can also do things like ‘in-painting’ of parts of an image to add or subtract things, kind of like a Photoshop content aware fill but 10 million times better.

Quite frankly im worried because I think it’s going to make most 2D artists irrelevant over night, and I think Open AI are going to make this an expensive commercial product similar to their GPT-3 that requires an expensive subscription and thus is generally used by companies to sell to other companies that want marketing or other similar articles written for them rather than hire a human to do it.

Maybe we as 3D artists don’t need to worry so much because this A.I. is not yet creating 3D objects? and we could simply use it to generate ideas for our 3D objects or scenes or animation??

What are your thoughts?

Extra video that shows a lot more examples of what it can do.

2

Every time a new AI comes out, we get a big long thread here about “oh no, artists are doomed!” and then a month later, everyone has forgotten about it.

Here’s the thing about AI: it can replace a lot of human work, repetition and pattern-detection specifically. It cannot have an original idea. AI is just another tool to integrate into an artistic toolbox. It can make art faster and easier but it can’t function without data sets to learn form, because it’s all just extremely advanced pattern recognition and replication. Without patterns to replicate, AI is helpless.

Let me give you a concrete example. There’s an AI copy writing software called Jasper (previously Jarvis). You give it a title of a blog, press Enter, and it gives you 500 words of grammatically correct text with facts, figures, sources, headings, etc. My wife uses this software everyday at work- she’s a content strategist working on a copywriting team. There’s the better part of 30 writers and strategists on this team. There’s actually more people now with Jasper than there were before.

Shouldn’t Jasper have replaced them all? Nope. It can’t. Jasper’s output is useless without human intervention. Because Jasper, like every AI, can’t actually have critical thought. It knows through pattern recognition what a “fact” or a “story” looks like, and it even knows how to write them correctly and make them human readable. The problem is that knowing the structure of a fact is NOT the same things as being able to distinguish between truth and falsehood (AI cannot do this). Jasper makes things up. It creates completely false statistics, using fake studies published in fake journals, to back these facts up. An AI can create the structure of a fact, but only a human can make sure it’s actually a fact.

Jasper, then, has become a tool this writing team uses. Not a replacement. Jasper can lay down the overall structure of a content piece, and then the writers go through and tweak it to be correct and better. As a result, a writer can whip out a content piece much faster, and so the company has way more clients, which means they can hire more writers.

To anyone concerned about being replaced by AI, I suggest you consider which parts of what you do are pattern recognition and replication, and which parts require the exclusively human abilities to generate original ideas and quantify data in an abstract way by concepts like “true” or “audience appropriate”. AI will help all of us work smarter and better, if we let it. But if you think AI can replace a human, you’re completely and entirely wrong. Just look at how horribly Tesla’s fully self-driving experiments are going- the number of accidents, especially involving emergency vehicles, is shockingly high (and easily findable with a quick Google search.)

6 Likes
3

So your contention is that A.I. can never reach human level in terms of creativity and you believe that there are exclusively human abilities namely generating original ideas and deciding if something is “audience appropriate”, and that AI can unquestionably never replace a human?

I’m sorry but i have to respectfully disagree, from the very latest tech I have seen seems to contradict your claims, and i’m certainly worried about the A.I. that will come along in the next few years say 1-5 years from now.

If you don’t mind me saying it seems that the first iterations of these A.I.s such as GPT-2, GPT-3 or DALLE-1 or VQGAN, or Tesla self driving software, have clouded your judgement about what A.I. can be capable of. It’s evolving fast and I personally think humans will be left in the dust in five years from now.

1 Like
4

You’re certainly welcome to think that if it makes you happy :slight_smile: I’ve seen a lot of technologies that were supposed to completely replace humans and yet somehow they never do.

I’ve yet to see any convincing evidence that AI is the magical outlier to this potential problem people have been worried about for centuries. You should know, this dates back to the Industrial Revolution, there were riots and protests and doomsayers galore when steam engines were first introduced. Huge protests also with the automatic loom.

They said the same thing about radio, photography, cars, trains, airplanes, computers, the internet, factory automation, CNC machines (specially for woodworking), and the printing press. “Humanity is doomed, the machines are taking over! We’re being replaced!” Still hasn’t happened, few hundred years later.

What’s especially interesting is that in all these discussions, the machines are never taking over just yet. It’s always right around the corner, just a few years away. You should really do some reading on the automatic loom, you’re using many of the arguments the anti-automation critics of the time used verbatim

5

This is an interesting viewpoint, but in my mind it can’t hold out for much longer, simply because I do not think there is anything mystical or supernatural about the human brain, I know that it is simply a complex, very energy efficient, biological neural network that controls a bio-chemical body. And thus it has hard limits on what it is capable of.

As A.I. and robotics evolve they will simply overtake humans and quite frankly leap far ahead of us very rapidly, and it might not even require sentience or self awareness, just a very good imitation of it that can easily fool a human.

And if it’s going to be harder to integrate A.I. into humans to augment their abilities than it is to just build an AGI controlled robot, then the AGI controlled robot is what we will create first. At that point there is no turning back, we will have lost, and AGI will rule and dominate the world.

1 Like
6

I disagree with you regarding the original idea part. In fact there are huge research efforts ongoing doing exactly this. One popular example is AlphaGo, AlphaZero, MuZero research projects which are built to come up with new patters and new ideas in games like Go and Chess. Thanks to that, those systems are able to progressively get better. They come up with completely new, never before seen ideas.
Even though most AI we are seeing today is based around pattern recognition (and replication as you mentioned), there has been a lot of ongoing and successful work which clearly goes beyond that. Usually those projects are also using the same kind of pattern recognition, but they also generate new training data in certain ways which introduce new “ideas”.

Edit: Nevertheless, I still don’t believe artists are going to be replaced! Many things will likely get simpler, the barrier of entry will be lowered, technical and tedious aspects are going to be reduced. Artists are still going to be able to produce art and we as humans still appreciate it.

1 Like
7

Please, peoples don’t say anything or fear without not know what is A.I. AI is special tree structure for programmers, that is it. Why you fear too much? Maybe you watch sci-fi film too much or read.

AI will only harm peoples from this sectors (maybe): Controllers, quality controllers, and some unskilled workers etc. That is it.

Creativity of the human is not only based their brain. We are not know what is consciousness, and we can not write algorithm about that. Without consciousness and emotions, you can not make art. Everybody can draw a paint, but can not everybody be an artist.

If you have a skilled profession, don’t worry.

2 Likes
8

Exactly. If you fear AI and don’t put in the effort to learn how to add it to your work, you’ll be replaced. If you learn how to use it as a tool, you’ll be a more effective artist, with better job options in the future. It’s that simple. If you get replaced by AI, that’s your choice- or rather, the consequence of your choice to not learn how to use the latest tools in your field

1 Like
9

Previously you stated “But if you think AI can replace a human, you’re completely and entirely wrong”, but now you state “If you fear AI and don’t put in the effort to learn how to add it to your work, you’ll be replaced”.

Now your contention is that a human working with A.I. as a tool will suffice to stave off being replaced in a job field.

But objectively this just sounds like humans desperate attempt to cling on and stay relevant and worth employing. When faced with the knowledge that A.I. is evolving at a rapid pace and that humans do not have something that is supernatural and that we do have hard limits on creativity and intelligence, just the same as any other animal with a brain ie a bird, a rat, a monkey etc, it’s pretty obvious to me that soon humans will not be needed, for every field there will be an AI that can work on its own more effectively than with a human guiding it. Does Alpha Zero need a human to guide it it’s moves? Not any more and Chess and Go were once considered special human domains that computers cannot best us at.

1 Like
10

AI can easily replace an ignorant human that refuses to move forward in life. It happens every day. As @hikmet said-

AI will only harm peoples from this sectors (maybe): Controllers, quality controllers, and some unskilled workers etc. That is it.

This isn’t an AI thing, though, anyone who refuses to make progress WILL be replaced. A 3D artist that refuses to use the latest software WILL be replaced, not by a magical AI robot, but by a younger and less stubborn 3D artist.

Here’s another concrete example: I work in web development, and I personally know people who used to work in web development but lost their jobs because they refused to get with the times. I know someone who was a dedicated Ruby on Rails dev who wasn’t interested in learning more modern technologies, like React frameworks, and they don’t work in web dev anymore. They can’t, because they aren’t hireable compared to more flexible people.

There’s TONS of AI in web development, I can point you to 100 different tools right now that can generate a website in just a couple clicks. None of them are used at a corporate level. Web development is one of the fastest growing fields right now. How can this be, if AI can do the work?

I’m not going to spell the answer to that out for you, since I’ve already said it a few times and Hikmet has as well. You are welcome to ignore that answer, but the “real-world” truth is still there. AI can write books and blogs, make websites, create images, write code, and more… and yet, all of those fields continue to grow, and continue to get paid for doing that work.

Here’s my final words on this subject:

  • AI is a tool, nothing more. It cannot do all the things humans can. It can do some things humans can’t.
  • Working with AI will make you more efficient and more hireable. Working without it, and living in fear of it, will leave you behind.
  • Every technology ever has had the same arguments around it- “humans are limited, the machines can do more! in a few years, humans will be obsolete!” Every. Technology. Ever. It’s never happened yet, and it never will
3 Likes
11
1 Like
12

So you are claiming there is no way that AI can learn what emotions are or what creates emotional reactions in humans to use that?

1 Like
13

Interview with Mo Gawdat, a happiness expert, former Chief Business Officer of Googe .
“8 years from now the smartest being on the planet will be a machine”. " by 2049 a machine will be a BILLION times smarter than us".

1 Like
14

Maybe reason is you only develop like mass product, without any creativity.

15

Smart? Computers already smart than human. But human is not equal smart creature. What makes human to human is consciousness, not intelligent.

We are Homo Spaiens Sapiens, means “person who knows what thinking.”

1 Like
16

For emotions, they must have consciousness. And we do not know what is “consciousness”. For this reason, never we can write any algorithm based consciousness.

17

I think i would agree that humans have consciousness and sentience and that makes us special in comparison to a machine. However to get back to the point of the post, I still fear that A.I. will replace humans or surpass humans in job fields, even without consciousness and even against humans that utilize A.I. as tools to augment their work.

1 Like
18

How they will that?

Personally I use AI based softwares for my job. If you don’t see yourself like mass product machine, then you have not any reason for fear.

1 Like
19

I fear DALLE-2 and DALLE-3 and DALLE-4. They can produce many 2D finished images for companies much faster than me and with very little guidance from a human, just one sentence, therefore the company does not need to employ an artist or pay for freelance work or purchase stock imagery, photography or digital art.

It’s that simple, companies will want to use it if its cheaper and can do a good enough job. It does not require solving the consciousness problem or anything insurmountable.

1 Like
20

Never.

You must think like that:

Example, if I am a client, and I want to work with pro artist. I have a dream, I want to tell you this dream and I want to you draw this dream. But I am not an artist. I try tell you what is my dream, but this is never enough. I write my dream and this software create a sketch. And you see this sketch and you can understand what I want and you draw this art with your pro skills, your emotions, your creativity and with add your own dream. Is this not good thing?