Ch-4 Generative AI

What is Generative AI?

●​ Generative artificial intelligence (AI) refers to the algorithms that generate new data that
resembles human-generated content, such as audio, code, images, text, simulations, and videos.
●​ This technology is trained with existing data and content, creating the potential for applications
such as natural language processing, computer vision, the metaverse(a virtual reality space), and
speech synthesis(generating spoken language by machine) .

Evolution of Generative AI

Early AI: Early AI were designed to recognize and categorize timings. These discriminative models could
be trained to identify objects or patterns, such as sorting emails into spam and non spam. These systems
could be categorized as supervised (training data with label) or unsupervised (training data without
labels).

Supervised Learning: This method involves training a model using labelled data. The classification of
data elements into categories or labels was initially taught to the machine learning models by humans.

Unsupervised Learning: In unsupervised or self-supervised learning, the machine learning model takes
unlabeled datasets and figures out patterns and inherent structures within them — without human
intervention.

Generative Modelling: Generative AI takes the unsupervised learning approach a step further by not
recognizing patterns but also creating new data based on what it has learned.
Timeline of Generative AI

Generative AI has evolved over several years to reach its current form. Over time, advancements in
neural networks and deep learning techniques have significantly enhanced its capabilities. From early
experiments in generative models to breakthroughs in natural language processing and image
generation, the development of generative AI has been a continuous journey of innovation and
refinement. Today, generative AI encompasses a wide range of applications, including text generation,
image synthesis, and creative content creation, showcasing the culmination of years of research and
development efforts.

Deep Fake: Deep Fake is a type of Generative AI that uses deep learning techniques to create realistic
and convincing fake audio, video or images. It often involves swapping faces, voices or other elements to
create a manipulated version of the original content. It can be used for entertainment, education or
artistic purposes, but they also raise concerns about misinformation, disinformation and potential for
malicious use.

Generative AI vs Conventional AI

Generative AI Conventional AI
Goal It creates new content It analyses, processes and
classify data
Training Generative AI models use vast Conventional AI employs fewer
libraries of samples to train complex
neural networks and other
complicated structures to
produce new content based on
those patterns.
Output Generative AI output is fresh, Conventional AI produces more
innovative, and often predictable output based on
unexpected. existing data.
Applications Generative AI benefits art, gaming, and design.
music, literature, gaming, and Conventional AI is used in
design. banking, healthcare, image
recognition, and language
processing.
Types of Generative AI

GANs (Generative Adversarial Networks)

GANs are neural networks that collaborate to produce fresh data Made up of two neural networks:
1.Generator Network & Discriminator Network

●​ The generator network produces the data, while the discriminator network analyses the data
and provides feedback.
●​ Until the generator can generate data that is identical to real data, the two networks collaborate
in a feedback loop.
●​ Examples-creating portraits of non-existing people, converting images from day to night,
generating images based on textual description, generating realistic video etc.

2.VAEs (Variational AutoEncoders)

●​ A Variational Autoencoder(VAE) is a type of deep learning model that consists of an encoder,

3.RNN(Recurrent Neural Network)

●​ RNNs are a special class of neural networks that excel at handling sequential data, like music or
text.
●​ They function by ingesting past inputs and applying that knowledge to anticipate future inputs.
●​ Example- Generating novel text in the style of a specific author or genre, predicting the next
character or word in a sequence etc.

4.Autoencoders

●​ These are Neural networks that have been trained to learn a compressed representation of data
●​ They function by compressing data first, then decompressing that compressed data to restore it
to its original form.
●​ Auto encoders can be applied to denoising or picture compression applications.
●​ Examples- artistic image creation, drug discovery. They generate highly realistic samples.

Examples of Generative AI :

Generative AI has many applications, from art and music to language and natural language processing.
Here are some examples of how generative AI is being used in various fields:

●​ Art: Generative AI is being used to create unique works of art.

For example, The Next Rembrandt project used data analysis and 3D printing to create a new
painting in the style of Rembrandt

●​ Music: Generative AI is being used to create new music, either by composing original pieces or
by remixing existing ones.

For example, AIVA is an AI composer that can create original pieces of music in various genres,
MusicLM by Google, musiclab.chromeexperiments.com/song-maker(creates your own song)
 ●​ Language: Generative AI is being used to generate new language, such as
chatbots(ChatGPT,Meta AI, Gemini, Co-pilot) that can hold conversations with users or natural
language generation systems that can produce written content.

Benefits of using Generative AI

Overall, generative AI offers a range of benefits, including increased creativity, efficiency, personalization,
exploration, accessibility, and scalability. By leveraging these benefits, businesses and organizations can
improve their content creation processes and provide better experiences for their users.

Limitations of Using Generative AI

●​ Data Bias: If generative AI is trained on biased or incomplete data, the output may be similarly
biased or flawed. This can lead to inaccurate or problematic results in certain applications, such
as in facial recognition or natural language processing.
●​ Uncertainty: Generative AI can produce unexpected and often unpredictable results, which can
be both a benefit and a drawback.
●​ Computational Demands: Generative AI requires significant computational resources to train
and generate its output, which can be expensive and time-consuming.
Generative AI tools

Here are some popular tools:

Artbreeder: Artbreeder is a web-based tool that enables users to generate new images by combining
different GAN models. Users can select and combine different GAN models to create new and unique
images.

Ethical considerations of using Generative AI

There are several ethical considerations that should be considered when using this technology:

●​ Ownership: There are questions about who owns the content generated by generative AI. This is
particularly relevant in creative fields such as music, literature, or art, where generative AI can
create original works that blur the lines between human and machine authorship.
●​ Human Agency: Generative AI raises questions about human agency and control. As technology
becomes more sophisticated, it may become increasingly difficult to distinguish between content
generated by humans and machines, which could lead to a loss of human autonomy and agency.
●​ Bias: Generative AI can replicate and amplify existing biases Bias present in the data used to
train the model. This can lead to harmful or discriminatory outcomes, especially if the generated
content is used in high-stakes applications such as hiring, loan approvals, or criminal justice.
 ●​ Misinformation: Generative AI can be used to create fake news or Privacy deepfakes, which can
be used to spread misinformation and manipulate public opinion. This can have serious
consequences for democracy and trust in institutions.
●​ Privacy: Generative AI can potentially be used to generate sensitive personal information, such
as credit card numbers, social security numbers, or medical records. This could be used for
malicious purposes.

The Potential Negative Impact on Society

●​ Generative AI can be used to create fake news or deep fakes that can spread misinformation and
manipulate public opinion.
●​ Lead to job displacement for humans who previously performed these tasks.
●​ Generative AI has the potential to generate sensitive personal information, such as social
security numbers or medical records, which could be used for malicious purposes.

Responsible Use of Generative AI

●​ Ensuring that the training data used are diverse and representative.
●​ The outputs are scrutinized for bias and misinformation.
●​ Prioritizing user privacy and consent,
●​ Having clear guidelines around ownership and attribution of generative content.
●​ Engaging in public discussions around the social and ethical implications of this technology to
ensure that it is developed and used in ways that are beneficial to society.