Home Tech & Gadgets Google claims that its text-to-image intelligence offers an “unprecedented photorealism”: what Imagen...

Google claims that its text-to-image intelligence offers an “unprecedented photorealism”: what Imagen is

25
0

Imagen is the company’s version of DALL-E from OpenAI, but is not available to the public.

cX.callQueue.push ([‘invoke’, function() {
googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-1567191399789-0’); });
}]);

Google has introduced an artificial intelligence system that can create text-based images. The idea is that users can enter any descriptive text and the AI ​​will turn it into an image. The company says the Imagen model, created by Brain Team of Google Research, offers “an unprecedented degree of photorealism and a deep level of understanding of language.”

It’s not the first time we’ve seen AI models like this. DALL-E from OpenAI (and its successor) generated titles as well as images because it is clever enough to turn text into images. Google’s version, however, is trying to create more realistic images.

cX.callQueue.push ([‘invoke’, function() {
googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-1567191803621-0’); });
}]);

To evaluate Imagen compared to other text-to-image models (including DALL-E 2, VQ-GAN + CLIP, and latent broadcast models), the researchers created a benchmark called DrawBench. This is a list of two hundred text requests that were entered in each template. Human evaluators were asked to rate each image.

They “prefer Imagen to other models, both in terms of sample quality and image-text alignment,” Google said.

Read:  SpaceX's Starlink satellite internet is now available for order in 32 countries: these regions may have their equipment shipped "immediately"

It is worth noting that the examples presented on the Imagen site are organized. As such, they can be the best of the best images that the model has created.

cX.callQueue.push ([‘invoke’, function() {
googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-1571296839761-0’); });
}]);

Like DALL-E, Imagen is not available to the public. Google does not believe that it is still suitable for use by the general public for a number of reasons. First of all, text-to-image models are usually driven by large data sets that are taken from the web and are not “cleaned”, which introduces a number of problems.

Google Image is not yet available to the general public

“Although this approach has allowed rapid algorithmic progress in recent years, data sets of this nature often reflect social stereotypes, oppressive views, and derogatory or otherwise harmful associations with marginalized identity groups,” the researchers wrote.

“While a subset of our training data was filtered to remove unwanted content, such as pornographic images and toxic language, we also used the LAION-400M dataset, which is known to contain a wide range of content. inappropriate, including pornographic images, racist insults and harmful social stereotypes ”.

As a result, they said, Imagen inherited “social prejudices and the limitations of large language models” and may describe “harmful stereotypes and representations.” The team said that the preliminary findings indicated that AI encodes social prejudices, including the tendency to create images of people with lighter skin tones and place them in certain stereotypical gender roles. In addition, researchers note that there is the potential for misuse if Imagen were made available to the public as it is now.

However, the team can allow the public to enter text in a version of the model to generate their own images: “In future work, we will explore a framework for responsible outsourcing that balances the value of external audit with the risks of unrestricted open access,” said the researchers. .

However, you can try Imagen on a limited basis. On its website, you can create a description using preselected expressions. Users can select whether the image should be a photo or an oil painting, the type of animal displayed, the clothing they wear, the action they take, and the setting.

So, if you’ve ever wanted to see an interpretation of an oil painting depicting a panda wearing sunglasses and a black leather jacket while skateboarding on a beach, here’s your chance.

cX.callQueue.push ([‘invoke’, function() {
googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-1571297016967-0’); });
}]);

Previous articleErdogan announces that Turkey will launch new military operations on the southern borders
Next articleLithuania wants the EU to have a fund for countries that receive Ukrainian refugees