Generate 3D Models Within 1-2 Minutes by Entering Text
Used for Game and Metaverse Production as Well as Design and Planning
[Asia Economy Reporter Yuri Choi] OpenAI, which gained attention with its conversational artificial intelligence (AI) 'Chat GPT,' has now unveiled a 3D model generation AI. This AI can create a 3D model in just 1 to 2 minutes by inputting text. It is expected that AI will expand its role from developers and researchers, as seen with Chat GPT, to architects and car designers.
3D Model Generation by Entering Words
According to the IT industry and foreign media on the 22nd, OpenAI recently released 'Point-E' on the world's largest open-source community, GitHub. Point-E is an AI that creates 3D models. By typing desired images in text, such as 'a Welsh Corgi wearing a red Santa hat' or 'a red mug filled with coffee,' it generates a 3D model. While OpenAI’s 'DALL·E 2,' introduced last April, was an AI that converted text into images, Point-E has evolved one step further as a 3D model generator.
Point-E itself is not a new technology. Google released 'DreamFields,' a 3D image generation AI, in 2021 and unveiled its successor model 'DreamFusion' this October. NVIDIA introduced 'GET3D,' a 3D image generation AI for the metaverse, in September this year. The difference lies in speed. According to OpenAI, NVIDIA’s single graphics processing unit (GPU) launched in 2017 can generate a 3D model within 1 to 2 minutes. It claims to be 600 times faster than Google’s DreamFusion. However, it still cannot render fine shapes or textures.
The rapid speed is due to the AI training method. First, like DALL·E 2, it learns text and images, then provides a 3D model combined with the image. Because it understands text, converts it into an image, and then transforms it into a related 3D model, the speed is fast.
Targeting the Metaverse Era, AI Directly Builds Virtual Worlds
OpenAI expects that once the model becomes more sophisticated, it can be used in game or animation production. This hints at the commercial use of on-demand 3D model generation. OpenAI explained, "Even people without professional 3D graphic skills will be able to easily create virtual worlds."
Experts predict that Point-E will have a significant impact on existing industries. It can be applied in various fields, from architectural structure design to interior design. However, concerns have been raised about ownership issues of the results created by Point-E, such as patent rights and copyrights.
Professor Kim Myung-joo of the Department of Information Security at Seoul Women’s University said, "AI is rapidly producing outputs not only in simple repetitive tasks but also in areas previously handled by experts," adding, "In the US and Europe, copyright is not being strictly enforced to promote AI industry development, but if this continues, the uniquely human domain will gradually narrow."
Meanwhile, as OpenAI rapidly introduces evolving AI models, Google is on high alert. According to The New York Times, Google issued a 'Code Red,' warning of a serious crisis regarding Chat GPT. Sundar Pichai, CEO of Google’s parent company Alphabet, instructed in a strategy meeting to address the threats Chat GPT poses to the search engine business. He also ordered the development of AI products similar to OpenAI’s image and graphic generation AI.
© The Asia Business Daily(www.asiae.co.kr). All rights reserved.
