Explore the Future Earth with Wander 2.0: AI Chatbot Driven By Knowledge-base Story Generation and Text-to-image Model

Explore the Future Earth with Wander 2.0: AI Chatbot Driven By Knowledge-base Story Generation and Text-to-image Model
复制标题

用 Wander 2.0 探索未来地球:由知识库故事生成和文本到图像模型驱动的 AI 聊天机器人

DOI:
10.1145/3544549.3583931
复制
发表时间:
2023
期刊:
Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems
影响因子:
--
通讯作者:
A. Asadipour
A. Asadipour
中科院分区:
--
文献类型:
--
作者:
Yuqian Sun;Ying Xu;Chenhang Cheng;Yihua Li;Chang Hee Lee;A. Asadipour

文献摘要

被引文献

相似文献

人们总是通过科幻(科幻)来畅想地球的未来,那么我们能否通过人工智能(AI)的镜头来创造一种独特的体验呢?我们介绍了WANDER 2.0,这是一个人工智能聊天机器人,通过在微信和不和谐等日常交流平台上基于知识的故事生成来共同创造科幻故事。WANDER利用谷歌地图中的位置信息,通过大规模语言模型(LLM)生成关于特定地点(如巴黎)的叙事性游记。此外,使用大规模文本到图像模型(LTGM)稳定扩散,WANDER传输与文本描述和现场照片匹配的未来场景,促进未来的想象力。该项目还包括在未来地图上实时可视化人类与人工智能的合作。通过与来自世界各地的游客的旅行,WANDER展示了人工智能如何作为连接虚构和现实的主观界面。我们的研究表明,多模式人工智能系统具有通过针对不同人自适应和独特的内容生成来扩展艺术体验和创意世界构建的潜力。Wander 2.0可在http://wander001.com/上获得
People always envision the future of earth through science fiction (Sci-fi), so can we create a unique experience of "visiting the future earth" through the lens of artificial intelligence (AI)? We introduce Wander 2.0, an AI chatbot that co-creates sci-fi stories through knowledge-based story generation on daily communication platforms like WeChat and Discord. Using location information from Google Maps, Wander generates narrative travelogues about specific locations (e.g. Paris) through a large-scale language model (LLM). Additionally, using the large-scale text-to-image model (LTGM) Stable Diffusion, Wander transfers future scenes that match both the text description and location photo, facilitating future imagination. The project also includes a real-time visualization of the human-AI collaborations on a future map. Through journeys with visitors from all over the world, Wander demonstrates how AI can serve as a subjective interface linking fiction and reality. Our research shows that multi-modal AI systems have the potential to extend the artistic experience and creative world-building through adaptive and unique content generation for different people. Wander 2.0 is available at http://wander001.com/