Practical No 3

GENERATIVE ARTIFICIAL INTELLIGENCE: Basic Definitions & Examples

Generative AI (genAI), broadly defined, refers to any type of artificial intelligence (AI) that can produce new text, images, video, or audio clips.  That it can create certain types of images, text, videos, and other media in response to prompts. Technically, this type of AI learns patterns from training data and generates new, unique outputs with the same statistical properties. Essentially, Generative Artificial Intelligence refers to AI models that have the ability to generate new content, such as text, images, or even videos, by learning from patterns and examples in large datasets. These models are designed to generate content that resembles what is found in the training data. It thus means that GenAI allows users to input a variety of prompts to generate new content, such as text, images, videos, sounds, code, 3D designs, and other media. It “learns” and is trained on documents and artifacts that already exist online

Generative AI models use prompts to guide content generation and use transfer learning to become more proficient. Generative AI evolves as it continues to train on more data. It operates on AI models and algorithms that are trained on large unlabeled data sets, which require complex math and lots of computing power to create. These data sets train the AI to predict outcomes in the same ways humans might act or create on their own. Early genAI models were built with specific data types and applications in mind. For example, Google’s DeepDream was designed to manipulate and enhance images. It can produce engaging, new visual effects, but the model’s development was primarily focused on image processing, and its capabilities do not apply to other types of data.

The field of generative AI is evolving quickly, however, and an increasing number of generative AI models are now multimodal. This advancement means the same model can handle different data prompts and generate different data types. 

For example, the same genAI model could be used to:

• Generate creative text

• Generate informational text

• Answer any type of question in a comprehensive and informative way

• Describe an image

• Generate a unique image based on a text prompt

• Translate text from one language to another

• Include the source of the model’s information in a response

Generative AI model development is often a collaborative effort requiring different research types, programming, user experience (UX), and machine learning operations (MLOps) expertise. A multidisciplinary approach helps ensure that generative AI models are designed, trained, deployed, and maintained ethically and responsibly.

Essentially, the relationship between artificial intelligence and generative AI is hierarchical. 

• AI refers to the development of computer systems that can perform tasks that previously required human intelligence. Typically, such tasks involve perception, logical reasoning, decision-making, and natural language understanding (NLU).  

• Machine learning is a subset of AI that focuses on discriminative tasks. It involves the development of algorithms that enable computers to make predictions or decisions based on data without being explicitly programmed how to do so. 

•  Generative AI is a subset of machine learning (ML) that focuses on creating new data samples that resemble real-world data. 

In addition, Generative AI refers to a subset of artificial intelligence that focuses on creating something new or generating new content rather than just analyzing or classifying existing data. Unlike other forms of AI that primarily deal with recognition or prediction tasks, generative AI models aim to produce outputs that are original and creative.

Generative AI models typically work by learning patterns and structures from large datasets There are various techniques and architectures used in generative AI, such as:

i). Autoencoders: These models are designed to learn efficient representations of the data by encoding it into a compact latent space and then decoding it back to the original form. They can be used for image generation, data compression, or anomaly detection.

ii). Variational Autoencoders (VAEs): VAEs are a form of autoencoder that not only learn the data's latent representation but also generate new data by sampling from the learned latent space. They are widely used for generating complex and diverse outputs, such as images, by learning the underlying distribution of the training data.

iii). Generative Adversarial Networks (GANs): GANs consist of two main components – a generator and a discriminator. The generator creates new samples, while the discriminator evaluates and provides feedback on the generated samples' authenticity. Both components continuously improve through competition, resulting in the generation of increasingly realistic and indistinguishable samples.

Generative AI has applications in various fields, including art, gaming, fashion, advertising, and even drug discovery. It enables the creation of new content, opens up possibilities for creative exploration, and contributes to the advancement of human creativity.

Some of the examples of Generative AI include the following:
a}. Text generation: Models like OpenAI's GPT-3 (Generative Pre-trained Transformer 3) can generate human-like text by predicting and generating words based on the given input. They can be used for various applications, such as generating articles, writing code, or even composing poetry.

b). Image generation: Deep learning models, such as Generative Adversarial Networks (GANs), can learn from a large dataset of images and generate new images that resemble the training set. They can be used to create realistic images, such as faces, objects, or scenes, or even generate artistic creations.

c). Music generation: AI models, like OpenAI's MuseNet, can generate new musical compositions by learning patterns from a vast collection of songs. These models can create music in various genres or even compose pieces with multiple instruments.

d). Video synthesis: AI models like Deepfake, based on GANs, can generate highly realistic videos by swapping faces or manipulating existing videos. While this technology has raised ethical concerns, it also has potential applications in areas like entertainment and special effects.

  Generative AI  has great potential to achieve much more  and can actuallly revolutionize various creative domains and academic research activities. .

ChatGPT, launched by OpenAI, became wildly popular overnight and galvanized public attention. (OpenAI’s DALL·E 2 tool similarly generates images from text in a related generative AI innovation.)

Gartner (n.d) sees generative AI becoming a general-purpose technology with an impact similar to that of the steam engine, electricity and the internet. The hype will subside as the reality of implementation sets in, but the impact of generative AI will grow as people and enterprises discover more innovative applications for the technology in daily work and life.

Within the fields of academic, the benefits of AI can include enhanced capacity for teaching and learning, research data processing and management, digital archival and text analysis among others. In the fields of commerce and industry, the benefits of generative AI may include faster product development, enhanced customer experience and improved employee productivity, but the specifics depend on the use case. Usually end users may determine the use to which they wish to put AI and what value they are looking to achieve, especially when using a service as is, which has major limitations.

Researcher and scholars should be careful in using AI because geneative AI  can create results and artifacts that can be inaccurate or biased, making human validation essential and potentially limiting the time it saves workers.

References

https://www.gartner.com/en/topics/generative-ai

https://www.techopedia.com/definition/34633/generative-ai

https://www.coursera.org/articles/what-is-generative-ai