By the XD Creative Engineering team, FCB Health New York, with the assistance of OpenAI's GPT-3 and ChatGPT
Generative AI, also known as Generative artificial intelligence (AI), is a type of machine learning that involves creating new content or data that’s based on certain input parameters. It can be done through techniques such as deep learning in which case a model is trained on a large dataset and can generate new content that’s similar to the training data.
One potential benefit of Generative AI is its ability to create content that is highly personalized and tailored to individual users. For example, a Generative AI system might be able to create personalized news articles for each reader based on their interests and browsing history. This type of AI could also be used to create customized marketing materials for businesses, such as personalized emails or social media posts.
Generative AI can create new content quickly and efficiently. Perhaps one of these models that’s trained on a dataset of music could generate new songs or melodies similar to the training data. This is useful for music production or for creating custom soundtracks for media projects.
Another use for this technology is in the creation of new, innovative ideas. For example, a Generative AI system could be trained on a large dataset of patents and research papers in a particular field and then generate new ideas for products or technologies. This type of AI is particularly useful in industries where there’s a need for constant innovation, such as technology or pharmaceuticals.
A diffusion model for protein design
Overall, Generative AI has the potential to revolutionize many industries and create new possibilities for personalization and innovation. Let's say a team has created a powerful new way to design proteins by combining structure prediction networks and generative diffusion models.
Generative AI has the potential to change the marketing industry in several ways. Here are a few examples of how generative AI could be used:
In the world of marketing, specifically for healthcare communications, Generative AI has a variety of uses as well.
For instance, a subset of generative AI includes text-to-creation. Using text-to-creation—specifically text-to-image programs—creative teams can utilize these tools when brainstorming and generating rapid-prototype ideas for storyboards, campaigns, videos, and much more. By using these tools, teams can significantly cut down the time needed to create drafts.
AI can also be used to create customized content for healthcare marketing campaigns. For instance, a healthcare company may use a Generative AI tool to create personalized social media posts for each of its target audience segments. The tool could analyze the characteristics and interests of each segment, and then generate content that’s tailored to their needs and preferences, including posts that address specific health concerns or offer solutions to common problems.
All in all, the use of Generative AI in healthcare marketing is another tool in the toolbox that can help teams more efficiently brainstorm ideas and help companies with their target audience more effectively.
What are large language models?
Large language models (LLMs) are machine learning models that have been trained on a very large dataset of text data and are capable of generating human-like text. These models are typically trained using a technique called transformer-based deep learning, which involves using multiple layers of neural networks to process the input data and generate output.
One of the main advantages of LLMs is their ability to be useful for a variety of tasks, such as machine translation, text summarization, and generating responses to questions or prompts. LLMs can also be fine-tuned for specific tasks or domains, such as medical language or legal language, to improve their performance.
For example, the Stanford Center for Research on Foundation Models (CRFM) and MosaicML announced the release of PubMed GPT, a purpose-built AI model trained to interpret biomedical language.
LLMs are being developed across wide range of industries. For example, GPT-3/ChatGPT developed by OpenAI and BERT developed by Google are two examples of such LLMs. They differ from each other mainly in terms of the size of their training datasets and the depth of their architectures. Life Architect showcases the various LLMs being developed.
Text-to-creation refers to the process of generating a creative output, such as a piece of artwork, a design, or a piece of music, based on input given in the form of a “prompt”. Text-to-creation is an umbrella term used to describe a variety of art created by AI, including images, video, audio, 3D objects, and stories.
There are various ways that text-to-creation can be used, including generating art or music based on a specific theme or idea, or creating a design based on a set of instructions. Some text-to-creation tools use AI to analyze the input text and generate a corresponding output, while others rely on preprogrammed templates or rules.
Text-to-image is a type of natural language processing task that involves generating an image from a given text description. The text description is in the form of a caption, a story, or even a paragraph that describes an image or a scene. Text-to-image generation can be used in a variety of applications, such as generating images for advertisements or brainstorms, creating illustrations for stories or articles, and generating personalized images for social media posts.
Current (popular) AI Text-to-Image platforms
Text-to-video, text-to-audio, and text-to-3D objects are all examples of natural language processing tasks that involve generating multimedia content from text descriptions.
Text-to-video involves generating a video from a given text description. Text-to-audio involves generating an audio file from a given text description, such as generating a voiceover or narration for a video. Text-to-3D objects involves generating a 3D model or scene from a given text description, such as generating a 3D model of a character or an object from a text description of its appearance and properties.
These tasks are all related to natural language generation, which involves generating text from structured data or from scratch. Text-to-video, text-to-audio, and text-to-3D objects are more advanced tasks that involve generating multimedia content from text descriptions, and they require more sophisticated natural language processing systems. These tasks are still being actively researched, and the quality of the generated content can vary.
You may wonder how genuine are the responses we receive from Generative AI technologies. The point of these technologies isn’t to create final products or perspectives, but to quickly iterate and help point users to ideas and answers that contain the expected content. Our recommendation is to use these technologies to inspire your own ideas and access answers that can significantly reduce the amount of time it takes to research.
Here are just a few examples of how we can apply Generative AI to our healthcare business:
Well, since you’ve made it to the end, we’d like to let you know something that you may or may not have already realized — this POV was almost entirely written by Generative AI through OpenAI’s Playground and ChatGPT. And all of the images seen here were generated by AI as well, using Midjourney. Pretty crazy, right?
Using text-to-speech AI from Eleven Labs, we created this recording to make the above content more universally acceptable and ADA compliant through voice. While we used the free version, Eleven Lab's paid version offers many more voices and languages. It's truly the most human-like text-to-speech synthesis we've heard so far.