OpenAI introduced a new version of the image generator DALL-E 3 – Kommersant

OpenAI announced 3rd version DALL-E neural network, which is capable of generating images based on a text description. It will be integrated into ChatGPT, will be able to better understand queries and generate more accurate and detailed images.

DALL-E 3 does a better job of depicting hands and can also accurately depict a scene with clearly defined objects and simulate how they will interact with each other in space. In combination with ChatGPT, the neural network will understand user requests even better and, if necessary, adjust the results. There will be restrictions on the use of images of public figures and “harmful prejudices” – this is how the developers want to protect themselves from the use of the neural network for propaganda purposes and for misinformation. Also, artificial intelligence will not copy the style of modern artists.

DALL-E 3 is currently in preview and will be available to ChatGPT Plus and Enterprise subscribers in October via API. Both DALL-E and ChatGPT come at a cost, while the only major free image generator currently available is Microsoft’s Bing Chat AI, powered by OpenAI’s GPT-4. The pictures received by users using DALL-E 3 will belong to the users. The company is also testing a new tool that will determine whether an image was created by a neural network or a real person.

Earlier American writers filed a class action lawsuit in Manhattan court against the company OpenAI, which developed the ChatGPT chatbot. They accuse OpenAI of copyright infringement, namely that the company used texts from writers to train ChatGPT illegally and without obtaining the necessary permission.

How AI slanders people and suffers from hallucinations, read in the Kommersant article “Bullshit Generator”.

Evgeniy Fedunenko