The world's largest image platform (the world's largest library prohibits uploading AI gen

1. Artificial Intelligence Image Robot

Artificial intelligence generates content, which is an indispensable supporting force for the future to fully move towards a new era of digital civilization. Recently, the world's largest library, Getty Images, announced a ban on uploading and selling illustrations generated using AI art tools such as DALL-E, Midjournal, and Stable Diffusion.

2. Artificial intelligence image generation

Getty Images CEO Craig Peters stated that the ban is due to concerns about the legality of AI generated content and the desire to protect website customers. There are real concerns about the copyright of these model outputs, as well as the unresolved rights issues of the images, image metadata, and individuals contained in the images, Craig Peters said.

3. High definition artificial intelligence image materials

Given these concerns, selling artificial intelligence artworks or illustrations may pose legal risks to Getty Images users, and we are actively seeking benefits for our clients. He added that in addition to Getty Images announcing a ban on uploading and selling AI generated images, websites such as Newgrounds, PurplePort, and FurAffinity have also made similar decisions. Why are AI generated content repeatedly banned? Is artificial intelligence generating content only harmful without any benefits? How great is the content generated by artificial intelligence? In 1950, Alan Turing proposed the famous "Turing test" in his paper "Computing Machinery and Intelligence", which provided an experimental method to determine whether a machine has "intelligence", that is, whether the machine can imitate human thinking to "generate" content and then interact with humans.

4. Artificial intelligence image creativity

To some extent, artificial intelligence has been expected to be used for content creation since then. After more than half a century of development, with the rapid accumulation of data, improved computing power, and enhanced algorithm effectiveness, today's artificial intelligence can not only interact with humans, but also engage in creative work such as writing, arranging music, painting, and video production.

5. Artificial intelligence image materials

In 2017, Microsoft's artificial intelligence girl "Xiaobing" released the world's first 100% AI created poetry collection "Sunshine Lost Glass Windows". In 2018, Nvidia released the StyleGAN model that can automatically generate images. It has now been upgraded to the fourth generation model StyleGAN-XL, which generates high-resolution images that are difficult for the human eye to distinguish between true and false.

6. Artificial intelligence image high-definition

In 2019, DeepMind released the DVD-GAN model to generate continuous videos, which stood out in clear scenes such as grasslands and squares. In 2021, OpenAI launched DALL-E and an upgraded version of DALL-E-2 a year later, mainly used for interactive content generation between text and images. Users only need to input brief descriptive text, and DALL-E-2 can create corresponding high-quality cartoon, realistic, abstract and other styles of painting.

7. Artificial Intelligence Picture Cartoon

It is reported that at the 2018 Christie's auction house, the world's first artificial intelligence artwork (painting) was sold for a staggering $432500. How does AI generated content work? The continuous iteration of artificial intelligence algorithms is the driving force for the development and progress of AIGC (Artificial Intelligence Generated Content). From the perspective of technological evolution, AIGC technology can be roughly divided into traditional template or rule-based pre deep learning stage and deep learning stage of rapid development of deep neural networks.

One of the methods for artificial intelligence images is to use object detection

In the early stage of deep learning, AIGC technology mainly relies on pre specified templates or rules for simple content production and output, which is far from flexible and realistic content generation. During this period, artificial intelligence algorithms did not have strong learning capabilities, but mostly relied on pre defined statistical models or expert systems to perform specific tasks. By cleverly designing rules, simple lines The generation of text and melody.

One of the methods for artificial intelligence film reading is to use object detection

For example, by defining a complex set of functional equations, the function curve drawn by the computer has some aesthetic pattern, but cannot pass the Turing test. The continuous iteration of deep neural network in learning paradigm and network structure has greatly improved the learning ability of AI algorithm, which is different from traditional AI algorithm. The loss function and gradient descent algorithm in deep learning can flexibly and quickly adjust the parameters in deep neural network, Thus achieving the learning function from data.

The process of artificial intelligence film reading is reflected in

Subsequently, with the introduction of new game learning paradigms and the emergence of generative adversarial networks, the ability of artificial intelligence to quickly apply content to different scenarios and tasks has also been greatly improved. In the future, with the continuous breakthroughs of deep neural network technology in both large model and multimodal directions, artificial intelligence generated content will not only be more realistic (actually challenging Turing testing), but also meet the requirements of diversity Advanced requirements for controllability.

For example, in the virtual digital world, the interaction between people, objects, and the environment is combined to form a whole scene. The combination of various elements in complex scenes, long texts, and other content cannot be separated from the technical support of multimodal large neural networks with super large scale and super multi parameter numbers. 03 Why worry about artificial intelligence generating content? Currently, although AI generated content has been able to achieve the level of authenticity, it has still been prohibited by institutions such as Getty Images, the world's largest image library, at the beginning of the article. Is the copyright issue they provide truly the core issue? Firstly, we can see that generating content through artificial intelligence is undoubtedly a huge productivity improvement for both media and operators.

Taking the latest multimodal large model DALL-E 2 proposed by OpenAI as an example, given an existing scene image, this model can add a specified target subject at a specific location. For example, if you say 'a tree is full of flowers', the algorithm can add different forms of fresh flowers at the specified location. When you need to adjust the content, you can also directly input' the tree needs to be closer to the upper left corner of the canvas', DALL-E 2 can automatically adjust the artwork, especially when you need to find an image that fits the theme of the article as the cover, artificial intelligence production of content is very time-saving and labor-saving.

We have also conducted copyright checks on images generated by artificial intelligence, and have not detected any copyright risks. This also means that the reasons of institutions such as Getty Images are not valid. In fact, there are only two real risks to the content produced by artificial intelligence. The first is that artificial intelligence algorithms have inherent defects, such as technical limitations that have not been overcome in transparency, robustness, bias, and discrimination, The direct result is that the content produced by artificial intelligence will conflict with ethics; The second challenge is the regulation of new technologies. With the continuous maturity of artificial intelligence technology, the content generated by machine deep learning becomes more realistic and can achieve the effect of "mixing fake with real".

Correspondingly, the application threshold is also constantly decreasing, and everyone can easily achieve "face changing", "voice changing", and even become a member of the "network water army". This is what should be worried about. Some of the viewpoints in the article come from the China Academy of Information and Communications and the JD Exploration and Research Institute END

Source: Techsoho