Business Implications of Using Generative AI to create and supplement dialogues in Entertainment media
We can help create the pipelines required to deploy the server-based approach and the device-based model. These are the possible marketplaces we have identified:
1. Leasing of server and LLM models in a PaaS (Platform as a Service) model, studios would need to pay a monthly lease or a contract for the time this platform is required
2. Leasing of LLM models and fine-tuning tools in a SaaS (Software as a Service) model, studios would receive the current production version of our tools to deploy on their own servers for a fee and pay a royalty based on the products usage
3. Leasing of finetuning pipelines and mobile models to studios which enables them to fine-tune and deploy mobile capable LLMs on their own for a royalty-based payment service.
Cloud Based Approach: We could deploy a large instance of a Large-Language-Model which runs on a server with multiple users and can utilise that power to cater to multiple players at the same time while only having to deal with a single server where all the logic is present.
This is especially suited for games which already run on online servers as all the studios have to do is add GPUs to their server as most game server loads are CPU based. If their server architecture is even a little modern, they need not do anything.
Local Device Based Approach: We could also create a pipeline to fine tune a small 1B or 1.5B parameter model and deploy it locally on each device, LLM’s already are small and concise enough to be able to run on mobile hardware. Leveraging this would be great as it allows the player to play the game without being connected to the internet and have solace in the fact that all data is secure in their own devices.
Tools-Technologies | WatsonX.ai, WatsonX.data |
College | |
Suja Mohandas' Comment
Generative AI is pretty good at predicting text which comes after a certain set of words, its what they are created to do, and it is what they do best. This makes them the best tool for this use case in the industry.
In all the years of production of media, the human element has been constant, a Storywriter or a team of storywriters creating new worlds for our enjoyment.
Our idea is to leverage the prediction and text creation power of generative AI to supercharge the art of storytelling. First a look at where the industry is at present
This is a screenshot from Skyrim, a game which came out in 2012. It was a cutting-edge game when it released but the technology in it can be considered pre historic compared to today’s tech. So what is current technology doing?
This is a screenshot from Cyberpunk 2077, a “Next-Gen” game from 2022, looks eerily similar right?
This is because the art of dialogue selection and choice-based story telling is limited by our time. We cannot in a reasonable amount of time account for the theoretically infinite number of choices made by players. The technology to implement it exists but humans do not have that much time.
But AI do, they are trained on more data than a single or even 10 people can hope to understand, given their training on our literature dating back hundreds of years.