Some say generative AI will allow anyone to create games and change the face of user-generated content. Popular tools such as ChatGPT have taken the tech world by storm, with many companies now looking toward AI to streamline their workflow.
Generative AI has the potential to be used as a tool during the development process of a game be it coding or the creation of imagery. However it can also be used outside of the creative process and aid with tasks such as advertising. Some are already embracing AI such as Ubisoft, whereas others are still sceptical.
In this guest post, Levi Matkins of LifeStreet shares his thoughts on generative AI and how it could help mobile advertising teams, what advantages AI presents and some of the issues to keep in mind.
What does the term generative AI mean?
Generative AI is a class of artificial intelligence that creates a variety of new and original content in the form of text, images, audio and video in response to prompts. This technology has gained traction recently through large language models (LLMs) such as ChatGPT and Dall-E 2. It differs from what we might consider “traditional” or discriminative AI systems in that text prompts create an output that previously didn’t exist before.
As an oversimplified example, with similar training data sets (lots of pictures of cats and dogs), a discriminative AI model can accurately determine whether a new image contains a cat or a dog. On the other hand, if you prompt a generative AI model to return a picture of a dog, it will create an entirely new image based on the images it was trained on.
Use cases for Generative AI in mobile advertising teams
The rise of these LLMs represents a huge shift in how people interface with AI models, and teams are now just learning to take advantage of this technology. There are obvious use cases for advertising teams already, including: summarising any content, generating keywords for SEO, headline and ad copy ideation, writing new blog posts, and any number of use cases for creating images.
How will these provide an advantage?
Overall, generative AI can provide a range of benefits and will massively reduce production costs of generating new text/image / and eventually video/audio content, improving cost efficiency and enabling personalised content at scale.
The implication is that content creation will soon be done for little to no cost if you use the right tools effectively. Using the right prompts will save teams time and money, and it's a skill that is currently necessary to generate optimal outputs. These AI systems are not (yet?) a replacement for human creativity or expertise but tools that can enhance or supplement certain skills in certain contexts.
The pitfalls of Generative AI
The biggest pitfalls in today’s models are referred to as ‘hallucinations’ – when the model very confidently ‘lies,’ returning completely false information. The system doesn’t have the ability to know that it’s hallucinating and if the user is unfamiliar with the content being generated, then that can lead to a lack of quality and accuracy. Fact-checking these language models is a critical component to effectively using them.
Future developments and how these can support advertisers
Even though GPT-3 was released two years ago, it took until the last few months for ChatGPT to be released and take the world by storm. GPT4 was recently released, and it’s safe to say we are still in the very early stages of learning ways to use these large language models to make them more valuable.
Right now, fine-tuning these models with domain-specific data is already happening and making an impact in specific industries. An existing example of this is GitHub’s CoPilot, which can greatly increase the impact of developers. Models will be fine-tuned on other datasets (of medical and legal documents for instance) for a range of new applications such as patient/client communication, legal review/document creation, and other novel use cases. In the mobile app world, app marketers will use generative AI to enhance and scale the creative asset development process. They’ll be able to dramatically lower production time for a single image and consequently the cost too.
We have to remember that the models and their outputs are just going to get better and better. While it’s too early to tell exactly what the future holds, we know this technology will be a major game-changer for many businesses.
Edited by Paige Cook