Can AI Make the Streaming Video Experience Even Better Than TV?

No, 天网不是准备在不久的将来消灭人类吗, 但人工智能(AI), 以及它的表亲机器学习(ML), are starting to make waves in industries across the globe. 从提醒医生注意潜在的, harmful pharmaceutical interactions to handling customer service inquiries to powering self-driving cars, AI is re-imagining what were once manually intensive tasks.

但这些技术并不新鲜. Pioneered in the ’60s and ’70s by such luminaries as Marvin Minsky and Douglas Lenat, humanity’s quest to create synthetic intelligence grounded in silicon and algorithms has been ongoing. 这只是最近几年的事, 虽然, that the technology has reached a state where it can be more easily applied to business processes that have become digitized. 在很多情况下, the application of AI and ML can lead to improved business efficiency, 节约成本, 甚至裁员. 但是这种技术对流媒体视频有什么用呢?

The adoption and proliferation of streaming video faces some serious challenges, 最显著的是规模和质量. 随着越来越多的人观看, 对高清甚至4K的期望, the greater strain on the delivery infrastructure (ultimately causing failures) and the more bandwidth consumed, all of which conspire to undermine the end-user experience. But what if video distributors could deploy computer intelligence such as AI that automatically makes decisions to improve the delivery and, 因此, 的经验?

Consider how much data is being generated by streaming servers and client players. This data could be automatically fed into AI systems, 哪些可以利用这些结果来改善交付, such as by rerouting to different CDNs when there is congestion. 但这并不能真正解决问题. There are still users who experience poor video quality while the switch is being made. What if that same system could proactively switch before failure even happens? This is the real power of AI—making operational decisions without input. Of course, the system naturally becomes smarter over time.

But video delivery isn’t the only place where AI might benefit. 以编码为例. Encoding vendors are trying to perfect video compression that considers the nature of the scene itself (i.e.环境敏感). 通过这样做, encoding time can be reduced (because not all of the scenes require the same level of compression) while still reducing the overall size of the finished video, thereby saving bandwidth for the end user without sacrificing quality. These context-aware encoders are backed by intelligent systems that may, 在某一时刻, even be able to rewrite encoding profiles to optimize compression without any human intervention.

And AI can be used to improve the video experience in other ways as well. 个性化和推荐呢? 内容发现呢?? 在这两种情况下, intelligent systems could learn so much about a user’s viewing behavior as to truly understand what the user likes to watch—but not just based on past data. 而不是, when combined with technologies like facial recognition (using a front-facing camera on a laptop, 例如), the system could learn to understand the emotional state of the user and, 当与其他数据(如一天中的时间)结合使用时, 天气, 日历活动, 等.), might suggest videos across platforms that the user would be most likely to enjoy watching. 当与视频中的交互性相结合时, this might lead to very personalized recommendations across ecommerce platforms like Amazon.

The future video experience is more than just “lean back.“它将整合社交和商业. 它将由数据驱动. And as distributors deploy AI systems (at the edge of the network, 当然, 以减少延迟), 随着时间的推移，哪个会变得更聪明, the gap between traditional TV and streaming will grow. 忘记天网. 认为TVNet. 看视频再也不会和以前一样了.

[This article appears in the September 2018 issue of 流媒体杂志 as "Can AI Make the Streaming Video Experience Better Than TV?"]