OpenAI is known for its various expertise in machine learning. In particular, she has created a team of bots that can play Dota 2 quite successfully and developed a tool capable of writing very believable but fake news. Another OpenAI achievement is the DALL-E program (a combination of “Wall-E” and “Dali”), which can generate images based on textual descriptions.
For example, if you enter “a cat made from sushi” or “a fox sitting in a field in winter,” you will get a set of corresponding pictures. The set of introductory words is still limited. However, the results are exciting. You can try DALL-E by this link.
DALL-E cannot only create pictures from several elements but also display them from different angles. Moreover, the program can add details that are not mentioned in the description but are necessary for a realistic image. For example, in the description for “a fox sitting in a field in winter,” the AI determined that a shadow should be added.
“Unlike the 3D rendering engine, whose inputs must be explicit and detailed, DALL-E can often fill in the blanks when the description implies that the image must contain a specific detail that is not explicitly specified,” the OpenAI team notes …
DALL-E has great potential with further development. He can understand how various objects change over time, capture the geographical context, create images in different styles, add missing parts of objects, etc. In the future, this can lead to various consequences, both good and bad. For example, such technology could end photographic stocks. OpenAI plans to analyze what ethical issues may arise with this technology and its impact on various areas of life.