AI Innovation

The use cases for generative artificial intelligence (AI) that have garnered the most interest in the past year have primarily revolved around how to use large language models (LLMs) to create various types of text. In the coming year, however, generative AI will also be more widely used to create videos.

Runway, for example, already provides a cloud service that has trained more than 30 AI models to create videos and graphics using existing content. Now the company is preparing to make available the ability to create videos and images using text-based prompts that almost anyone can master in much the same way DALL-E enables them to create art using text-based prompts.

Use cases for the Runway generative AI platform includes everything from embedding videos in sales presentations to reducing the cost of making films. “We’re just scratching the surface,” says Michelle Kwon, head of operations and partnerships for Runway.

Pricing for the Runway service starts at $.05 per second of generated video and Runway is moving to integrate its service with the online graphic design tools provided by Canva as part of an effort to make the platform more accessible. In addition, Runway has partnered with Getty Images to provide its enterprise customers with a way to legally use the library of images that Getty Image curates to create videos. Getty Images has also launched its own generative AI service for creating images.

The only caveat is that Runway is putting guardrails in place to make sure anyone using the platform is only using intellectual property (IP) they either own or have permission to use to generate videos, notes Kwon.

There are, of course, similar services being offered by OpenAI, Adobe and others. The primary issue is that, given copyright issues that remain largely unresolved, organizations will need to tread carefully. Getty Images, for example, is suing multiple providers of generative AI services for creating images that were trained using IP that belongs to the company. Much like the case that the New York Times is now pressing against Microsoft and OpenAI, the argument will come down to what constitutes fair use of IP versus copyright infringement. Providers of generative AI platforms contend that content that is publicly available on the Web that is used to train an AI model to create new content does not infringe copyrights. However, the New York Times, along with a number of authors that have launched copyright infringement cases, have been able to cite instances where the output generated by generative AI platform such as ChatGPT is the same as the copyrighted content they published.

Microsoft, OpenAI, Google and Amazon, in response to these legal challenges, have all promised to indemnify customers from any lawsuit filed. However, while an organization may not have to bear the cost of the lawsuit, they will, if they lose the case, need to remove content that is deemed to have violated a copyright. For organizations that create content that is commercially sold, the risk to any revenue stream that might be impacted remains significant. It also raises the question of the financial stability of any generative AI platform that might be required to pay billions of dollars to license content to train an AI model.

Of course, there is still plenty of free content that can be used to train multi-model generative AI platforms capable of generating text, images and video. The issue is that copyrighted content tends to be of a higher quality, simply because the individuals that created it had the necessary financial motivation. There may come a day when providers of generative AI services wind up paying individuals to create content to train AI models, but until then the current level of uncertainty involving IP is likely to create some cause for pause within larger enterprises that are typically not eager to create any more additional legal jeopardy than they already have today.