By Patrick Trivuncevic
Artificial Intelligence, A.I., is a hot topic in many industries with Media and Entertainment poised to embrace more intelligent data management and automation in the coming years. Media companies have lead the explosive growth of big data technologies because it enabled them to drive digital revolution, exploiting more fully not only data which was already available but also new sources of data such as content from social media sources and user generated content. Media companies are increasingly looking to generate revenue from delivering content and data through diverse platforms and products. Increased and better integration of solutions along the data/content value chain will be fundamental in order to convince decision-makers to invest in new technologies such as Artificial Intelligence.
What is AI in Media and Entertainment?
AI is a vast field that covers a gamut of loosely related technologies. Colloquially, the term "Artificial Intelligence" is applied when a machine mimics "cognitive" functions such as "learning" and "problem solving" and has the ability to make decisions based on past experience. Differing from simply executing a fixed set of procedures and algorithms that have been pre-programmed as software, AI systems are able to “learn” about actions and results and adapt their behaviour to optimise outcomes and deal with mutable inputs. Cognitive analytics and Machine learning are closely related to AI.
Taking a look at a recent AI Survey in Media sponsored by Quantum in partnership with TVTechnology, There’s Nothing Artificial About AI for Broadcasters, we find that AI is emerging at a rapid rate with 21% of survey respondents having some experience or adoption today. A further breakdown of 21% who have had AI exposure fall in a few key categories.
How organisations with AI experience are currently using the technology
At the top, as we may expect, is Automated metadata creation. Media companies with any content library or repository, particularly those with large content libraries have always faced challenges with meaningful automated metadata creation. Solutions have existed for a while that are able extract metadata from files or file types, providing information such as date and time, format, codec information, etc. and that allow for additional custom meta data to be embedded or supplemented. Custom metadata, which is often more useful than date and time or format and codec metadata, is only effective if it’s applied by some process in the first place. Unfortunately, many solutions to date are not able to automatically define and describe unknown content. Custom metadata is often a manual process, often requiring a dedicated human resource(s) to enter the custom metadata. It is further reduced in value and meaning when the custom metadata is templated and semi-automated with customisation generalised or normalised to categories to simplify implementation and automate execution.
Common deficiencies connected with limited metadata
Automating custom metadata with AI is unsurprisingly the starting point for tackling deficiencies in meaningful metadata. If we look further into what custom metadata creation is possible we find the AI applications able to do facial recognition, scene detection, sentiment, landmark recognition, object recognition, logo recognition and translation. Vendors such as IBM (Watson), Veritone – Quantum (AiWare) and the various major cloud providers Google/Alphabet, Microsoft and Amazon all have solutions and partners employing AI technologies for meaningful custom metadata creation. MAM Vendors are another obvious choice for AI adoption with many beginning with better search/Intelligent search. For better search or intelligent search vendors like Empress, who are a MAM Vendor, have a need to generate or gather metadata and have facial recognition and transcription options to provide the data for the intelligent search.
Metadata creation is only one of many applications for AI in Media. Applications such as automated captioning, quality assurance, and automated clip generation and distribution are just some of use cases for AI.
AI ‘in’ Storage and Data Management
People often think cloud or big companies like Google or Amazon when they think AI and for good reason. Big companies like Google or Amazon have seemingly limitless on-demand computing resources and those companies have been some of the earliest adopters and pioneers of AI technologies. IBM’s Watson and Alphabet (formerly Google)’s Deep Mind all have their place in the AI story but upcoming start-ups and established companies not normally synonymous with AI are finding their place amongst the titans of AI. Quantum a well-established Storage and Data Management Vendor has forged a strategic relationship with Veritone adopting a version of AiWare for its Xcellis workflow Storage architectures. Quantum has a large customer base with 100’s of Terabytes to many Petabytes of data on various storage technologies. A lot of the data is archived or historical, this presents a challenge and opportunity in itself. What if your data management tools or storage including archives could be analysed with AI? How does one go about that?
One of the challenges with AI adoption in Media is oddly enough the cloud requirement from titans in AI. There is no question everything is moving to the cloud but moving large amounts of media to the cloud today remains a challenge. AI engines in the cloud often require the data gathering to be performed in the cloud. This presents a problem with many networks in geographies around the world having poor connectivity and bandwidth. Content from media organisations are measured in TBs and PBs today, uploading this content is often not possible in adequate timeframes. In many cases content is not allowed to exist outside of certain locations such as on premises or certain geographies.
Recognising this challenge for many of its customers Quantum’s strategic approach with Veritone’s AiWare allows for on premise data gathering and cognitive analytics.
AiWare can be deployed on certain Quantum Xcellis Workflow Extenders or alternatively an appropriate appliance can be supplied. Content and metadata can then be processed and analysed through the storage data pipeline. Online, Near-line and Archived data can be analysed bringing new search capabilities and consequently additional value to the customers content. Customers with or without a MAM will also benefit from metadata rich content from the creative user searching for the content to organisations wanting additional means to monetise content.
Monetisation and Marketing, often related or hand in hand is another area of any business that could benefit from the cognitive analytics that can be gleaned from existing content or data.
Insights from customers data could provide high performance and highly accurate results in key metrics for delivering competitive advantages in the fast-paced Media and Marketing sectors.
The graph below shows the various other categories AI is currently beginning used.
Respondents that currently use or have used AI technology (by category)
With more Vendors expected to adopt some form of AI, machine learning or cognitive analytics into their solution offerings and products in the coming months and future, we can only begin to imagine the possibilities and opportunities AI will deliver.
If you’d like to know more about how Artificial Intelligence could help your business automate workflows and/or maximise your content's value, please contact us at Digistor.