Nice, fellow scholars, this is too many papers with doctor kind of I hate early, twenty nineteen.
The learning based technique appeared that could perform common natural language processing operations, for instance, answering questions, completing text, reading, comprehension, samurai’s asian and more this mess that was developed by scientists at open ai and they called it g p t to the key idea for a g, p d: two was that all of these problems could be formulated as different variants of text.
Completion of problems we’re all we need to do is provided an incomplete piece of text and it would try to finish it. Then. In june, twenty twenty kg p t three, that’s supercharged this idea and among many incredible examples it will generate website layouts for my written description.
However, new one sad that these neural networks can only deal with text, information and sure enough, a few months later, scientists an open any. I thought that if we can complete text sentences, why not try to complete images?
Do they call this project image g p t and the problem statement was simple: we give it an incomplete image and we ask the ai to fill in the missing pixels. It’s good identify that to get here likely hold a piece of paper and finished.
The picture accordingly hand even understood that if we have a droplet here- and we see just a portion of the ripples, then this means a splash must be filled in and now right in january, twenty twenty one jeff, seven months after the release of g p d.
Three here is their new mind, blowing technique that explores the connection between text and images, not finishing images, already kind of works. So what’s new thing, can they do in just a few moments? You will see that the more appropriate and would be what can they do for now?
Well, each creates images for my reading text captions and do as in a moment how monumental of a challenge that is the name of this technique is a mix of salvador, dali and big sars wally. So please meet dolly, and now, let’s see see it through an example for neural network based learning methods, it is easy to recognize that this text says open, yeah I and what a store front as images of both of these exist in abundance.
Understanding that is simple, however, generating the store front that says, open ai is quite a challenge. Is it really possible that it can do that? Well, let’s try it look. It works wow. Now, of course, if you look here, you immediately see that it is by no means perfect, but let’s marvel at the fact that we can get all kinds of to the and three tax milk at storefronts from different orientations, and it can deal with all of these cases reasonably well and of course it is not limited to store fronts. We can request, license plates bags of chips, neon signs and more.
You can really do all that. So what else well get this? It can also kind of invent new things. So, let’s split our entrepreneurial hat on and tried to invent something here, for instance, nest tried to create the triangular lock or pentagon, or you know just make it the hexagon.
It really doesn’t matter because we can ask for absolutely anything and get a bunch of prototypes in a matter of seconds. Now, let’s make it wide and look now we have a happy happy kind of way. Why is that? It is because I am the night transport researcher by trade.
So the first thing I look at when seeing these generated images is how physically plausible they are, for instance, look at the sweat luckier on the blue table and it not only put it on the table, but it also made sure to generate appropriate glossary of elections that matches the color of the clock.
He’s can do this to loving it. Apparently it understand geometry, shapes and materials. I wonder what else does you understand? Well get this, for instance, even understand styles and rendering techniques being a graphics person.
I am so so happy to see that he’s learned. The concept of flu polygon counts, rendering isometric views, clay objects and we can even add an x ray view to the isle kind of and now, if all that wasn’t enough hold onto your papers, because we can.
