How AI is Disrupting Video

IBM’s new publication, featuring research from Gartner, examines how enterprises are leveraging unstructured video data through machine learning. Topics, showing how AI is disrupting video, range from video analysis for business value, AI powered transcription and examining a large scale deployment that uses cognitive capabilities to enhance video assets.

Check out the publication on AI and streaming video, which includes links to full papers and videos on these topics. Below is a quick recap of some of the information touched on within.

Getting started with image and video analysis in your organization

By 2022, 80% of video/image analysis projects will involve some form of deep-learning model. AI analysis and utilizing artificial intelligence on previously “dark” data, such as image and video assets versus written content, presents exciting opportunities. In fact, the ability to exploit the value of large unstructured image data is a huge untapped opportunity that faces organizations today.

Enterprises, though, need a way to determine that investments in this technology is offering clear business value. Included in the newsletter is a guide, Seven Steps to Get Started With Image and Video Analysis in Your Organization, by Gartner that examines this topic in more depth. It includes recommendations, examining risks and covering how to begin using this technology.

Read the publication on AI and streaming video for more details.


The future of video communications

Also included in the newsletter is how IBM is powering enterprise video communication while leveraging AI. This is being used to enhance the end user experience and also improve asset discovery. Through using Watson’s artificial intelligence capabilities, enterprises can use IBM’s solutions to generate transcripts and captions through converting video speech to text.

This process is used to provide captions for assets, from a list of supported languages. This can aid in increasing understanding of content while also expanding the potential audience for those who may be hard of hearing or deaf. After being generated, these captions can be edited for accuracy or to match personal preference in structure.

How AI is Disrupting Video

This transcription is not only used to add captions, though, but also to make it easier to find relevant videos and even relevant moments within them. This is achieved through making these transcripts and captions searchable. This can be done on a portal basis for enterprise video search and discoverability, honing in on the appropriate assets. However, it can also be done to jump to the exact moment of relevancy. This can be incredibly valuable for long form content, negating previous experiences of having to watch hours of content just to search for a specific fact. It can also make large archives of content more accessible and useful through giving end users tools to search and find needed information.


IBM CIO office and scalability

Included in the newsletter is also how IBM itself is using this technology on a large scale deployment, servicing over 350,000 employees with live and on-demand video in over 170 countries. This gives employees a way to quickly find assets, from an infrastructure that has been built around the goals of reliability and scalability.

The scalability component has been achieved through using SD-CDN (Software Defined Content Delivery Network) technology. This utilizes multiple CDNs (Content Delivery Networks) for both increased global reach and an improved end user experience. On the topic of global reach, a CDN ideally has server presences on a global scale, able to decrease the physical distance between an end user and a server. However, certain CDNs are stronger in certain parts of the world, and by using multiple CDNs viewers can get improved coverage.

In terms of an enhanced end user experience, this is achieved through built-in QoS (Quality of Service) elements in the player. Essentially, the player will send back playback performance data. In doing so, if it notices that an abnormal amount of buffering is happening it will switch the viewer to a different source. This can be a different edge server within the same CDN or an entirely different CDN. The actual switch is virtually seamless, similar to switching between bitrates through the adaptive bitrate process. Through this approach not only can a viewer get a better experience, but also it can mitigate the possibility of an outage. For example, the abnormal buffering could have been an early warning sign. A hint of a systemic issue with the edge server or CDN. As a result, by moving the viewer to a different delivery method it might have avoided the problem before it really started.

In addition to traditional CDNs, SD-CDN can also work with IBM’s Enterprise Content Delivery Network (ECDN). Now ECDN is a virtual appliance that can be installed on shared or dedicated hardware. It’s a virtual server that can work in conjunction with firewalls and includes flexibility for multiple instances to support multiple ISPs or multiple offices. Once installed, ECDN caches assets to mitigate strain on the local network. For example, rather than having 200 people in an office all using 2mbps each (total of 400 mbps) for an all hands meeting, over taxing the available download speed, that usage can be reduced to just a few. In other words, total consumption might be dropped to just 3 or 4 mbps, dependent on available bitrates that are being cached. This solution works with SD-CDN as well. For example, if a viewer goes “out of range”, like they start watching the content from the office on their phone through ECDN and leave, they can be shifted toward a traditional CDN to keep watching.

Read more on IBM’s SD-CDN technology in this white paper on Scaling Video Delivery to Reach Massive Audiences. For more details on how the IBM CIO office is delivering video at scale internally, also check out this archived How to Scale Your Corporate Communications, which is also mentioned in the publication on AI and streaming video.



The newsletter provides a number of assets that dive deeper into topics around how AI analysis is improving video for businesses. These range from image analysis to audio analysis, both to improve the insight on assets but also enhance how they are used by viewers as well.

Looking for more information on delivering video internally and what determines a successful deployment? Be sure to check out our Using Video for Internal Corporate Communications, Training & Compliance white paper. This covers use cases for internal video while also examining ways to improve per asset ROI.