What is Generative AI and how it differs from traditional AI/ML techniques?
Introduction:
Generative AI is subset of Artificial Intelligence who has deepen its root over past few years. Gen AI is a term used to describe the work or output that is produced artificially with help of AI techniques without actually intervening the system. Gen AI is widely used term in recent years as it provides advantage over traditional systems.
What is GENERATIVE AI?
Generative AI is system that is capable of producing new and original content on providing some prompt. These systems are built over massive dataset, learning patterns and relationships in data. Generative AI models once trained can generate various types and forms of output as needed.
Types of Gen AI:
Generative AI can be classified as per the results they produce. Majorly these models are classified as below:
• Text Generation: In this type of model the system generates text on basis of the prompt given. This text may be either analyzed or trained part of the system. Model is capable to generate text on its own using algorithms and trained AI models. For example, on providing prompt: “write a creative 1 line in Shakespeare's style of writing on "situation of today's world". Following text is generated: “A world awash with turmoil, where hope and fear do intertwine.”
Example for text generation: Open AI’s GPT-4, Google’s Gemini/Bard, Microsoft’s Copilot, etc.
• Image Generation: This model is capable to generate or illustrate an image according to given prompts. For example, if a prompt is given “ Generate an image of Human habitat living and advancing on planet Neptune”. Following image is given as output
Example of Image Generation: Dall-E, Midjourney.
• Audio Generation: These models are widely used for dubbing videos or to generate animated movies. These models give further a softer hand to tune the audio in whatever domain we need. They generate audios using the phonetics of the text provided. The models are trained on actual voice of humans of various dialects, speech rate and spatial uniqueness.
Example of Audio Generation: Tacotron, Whisper, Wavnet, etc.
• Video Generation: Video Generation is an advanced technique to generate video on basis of single or series of prompts. These models are currently in under developing stage but are capable to produce high end outputs.
Example of Video Generation: Imagen, Stable Diffusion Video, VQ-VAE-2
Difference between traditional AI/ML and generative AI:
Goal: GenAI is concentrated on generative tasks, whereas traditional AI/ML is mainly used for predictive tasks.
Output: While GenAI generates fresh, original content, traditional AI/ML makes predictions and insights based on data that already exists.
Applications: The field of generative AI is more diverse and includes fields like materials science, drug development, and content creation.
Conclusion:
Generative AI is a major advancement in the field of artificial intelligence. Its capacity to produce fresh, original content has the capability to transform whole markets as well as our way of living and working. We can anticipate a future full of fascinating and cutting-edge possibilities as GenAI develops further.
