Making sense of AI: Grabyo’s 2024 predictions

By Gareth Capon, Grabyo CEO.

While 2023 has been a challenging year for the media industry, it has been a catalyst for change.

With economic uncertainty leading to market consolidation and tightening of budgets, media teams have been tasked with maximising efficiency while creating more content and enhancing the fan experience. Not an easy feat.

Broadcasters, federations, leagues and teams are using technology to accelerate, automate and monetise output, with a greater focus on streaming, which now dominates media consumption for almost every age group.

Yet streaming is not the biggest media tech story this year. Two topics have dominated conversation in our industry in 2023: cloud services and AI.

AI goes mainstream

AI has been used in sports technology services for years, but everything changed with the launch of ChatGPT. ChatGPT was the first mainstream consumer-facing service built on a large-language model (LLM) and it evolved rapidly to more than 100 million weekly users. Other models followed, including Bard from Google (Alphabet) and a range of AI-services from AWS in AWS Bedrock.

The space is developing incredibly quickly, culminating with the launch of Custom GPTs from OpenAI in November. Custom GPTs make it easy for anyone to design and customise their own AI model, utilising the underlying capabilities of ChatGPT. In just a couple of weeks, hundreds of new Custom GPTs emerged, covering everything from avatar generation to AI-based sports commentary. We will see many more sports-specific GPTs as we move into 2024.

Other generative AI models are able to produce images, including video frames, either using a visual reference or from scratch. This includes DALL-E, RunwayML and Midjourney as well as established providers such as Adobe, which is building AI into its core products with the launch of Firefly for which users can extend photographs and vector images with a simple text prompt within the software.

What does this mean for live broadcasting and video in sport? The most exciting applications of generative AI and LLMs lie in the contextual understanding of video, metadata services and content augmentation.

Automation and metadata

Automating the creation, contextual understanding and management of video metadata using AI is a focus for the sports industry.

Creating useful metadata without hours of manual input has clear benefits to rights owners and publishers, especially for those managing high volumes of content or a huge sports media archive.

Companies such as Magnifi.AI are leading the way in using computer vision to analyse video, audio and text from sports broadcasts and archives to identify events and create valuable metadata. There are a number of use cases for this, such as automating markers in recordings of broadcasts to identify key moments – goals scored in football, sixes hit in cricket, an ace in tennis. For real time video, highlights creation and post production workflows, this significantly reduces the effort and hours required by editors. It also includes adding contextual metadata to a video archive using facial recognition, audio analysis and emotion detection, moving from the identification of objects to the understanding of scenes, action and flow.

This also applies to real-time workflows for live clipping – social media teams are able to scale their output, assisted by AI, but retain storytelling control.

Then you have further useful metadata, such as logo recognition and face recognition, all generated behind the scenes, logging away content with ultra-specific descriptions for later retrieval. This would require a large team and many hours to achieve manually.

This technology is already in use by many sports organisations across the world. However, the challenge for providers of this technology is offering the specificity required to create useful metadata for each and every sport – but this is coming. The end game is to personalise output on an individual basis, combining live sports coverage with the interest graph of an individual fan. Combined with developments in data, alternate broadcasts and interactivity will create an entirely new model for sports media and OTT.

LLMs and metadata

Building on LLM foundations, many applications are being developed that harness their power to automate manual workflows.

Using a MAM system optimised with metadata, you could wrap the functionality of an LLM to improve retrieval and search of databases.

Those with large content libraries could vastly improve the flow of assets across their content supply chain. Imagine automated metadata insertion coupled with advanced search in the style of ChatGPT – production, digital, social and marketing teams across organisations could have their entire content library shared, organised and at their fingertips. This is much more than identifying an object in a video, but a move to understand what the video represents, why it may be important and how it fits in the context of a wider story or search prompt. Eventually this will be a proactive process, whereby a GenAI model can suggest content to use, rather than being asked to find it.

Remaining human-centred

Many of us are excited and anticipating the peak of AI to reduce and remove manual tasks, but it is clear that AI workflows are most powerful when, at their core, they’re designed for humans.

A great example of this is using AI for captions and speech-to-text services for live broadcasts. Not all sports have the capabilities or budget to generate live captions for every video broadcast, using AI to create this text in real time significantly reduces the cost of this process while ensuring the video is accessible for everyone.

We’re in the storytelling business, and we must remind ourselves that while AI can do many things, it will be some time before it can replicate the capabilities of human editors to craft stories that capture and elicit emotion, that is if it ever can.

Cloud transformation

2024 will see a significant shift towards cloud transformation in live production across the broadcast, sports and media sectors. This is driven by the increasing decline in linear TV, loss of TV ad revenue and a shift to streaming across OTT platforms, FAST channels as well as YouTube and social media.

The need to cut operating costs and adopt sustainable production practices is accelerating this transition. The adoption of cloud-based technologies is seen as a key trend for the media industry in 2024, offering greater flexibility, scalability and cost-efficiency than legacy systems. This shift is not just a technological upgrade but a strategic response to the changing consumption patterns and production needs in the media landscape, underscoring cloud technology’s role in shaping the future of live sports and broadcasting.