2023年4月17日

酉-启动Startacus让互联网安全与人工智能的管理创新在线内容适度;统一是一个创业公司,开发了上下文感知的人工智能检测有害内容和保持品牌和平台安全。有人勇敢地看一眼YouTube视频的评论就知道有多少垃圾被发现。无论是小youtube用户点击试图挑起战争,无关政治呀呀学语,或机器人,有很多废话,很多是非常不合适的。在一个流聊天也是一样。自动节制这些正在逐渐改善,但仍然会通过很多。这突出了另一个问题:如果审核的评论是如此困难,我们如何缓和80 -年的价值的视频内容上传到互联网的每一天?答案显然不是人类的节制。唯一现实的方法,以确保这些内容是免费的从可怕的主观brand-unsuitable是采用人工智能。伦敦启动酉是建立一个强大的人工智能,能够处理超过每秒25000帧的视频,缓和与启动声称是人类的精度水平。 It can do this by taking into account crucial details from what surrounds the video, not just what it is seeing in those frames. This includes things like the title, captions, information boxes, comments, and generally as much context as it can see. One example the startup uses is that nudity in a medical video is very different to nudity in other contexts, and the AI should be capable of determining that, just as it should be able to tell the difference between glorifying drug-taking and demonstrating how to use an EpiPen. Current automated systems not only miss a lot, but they can also do a lot of harm to creators whose videos are demonetised or taken down entirely because of the system falsely determining them to be unsuitable, and even if an appeal is successful, the damage is already done - that creator has lost revenue and the view spike their video would have received when it was new. Early this year, Unitary announced that they had raised £6.7 million in investment from the Plural investment fund and additional angel investors. With the amount of online content only increasing, this funding will boost this much-needed intervention in moderation. Subscribe to our newsletter If you would like to receive our startup themed newsletter, full of the latest startup opportunities, events, news, stories, tips and advice, then sign up here. <