Skip to main content

Discriminative and Generative AI

Discriminative Artificial Intelligence(AI)


Discriminative Artificial Intelligence(AI) models are taught to distinguish between different data classes, patterns, images etc. In simple words it is a type of artificial intelligence uses Machine and Deep Learning techniques that helps us tell different things apart. Discriminative AI focuses on learning the lines that separate different groups in our data. Instead of making new data, it looks at data we already have to figure out what it is.

• Application Areas

a. Image Recognition: Discriminative AI helps identify what's in pictures, like telling if it's a cat or a dog.

b. Speech Recognition: It figures out different words or phrases when people talk.

c. Natural Language Processing (NLP): It sorts text into different groups, like figuring out if a review is positive or negative.

  • Working of Discriminative AI

Discriminative AI learns from examples, like pictures of cats and dogs, to get better at telling them apart. It figures out the lines that separate different things, and the more examples it sees, the better it gets. When given new data, it can tell what it is based on what it learned before.

As technology gets better, discriminative AI will likely get better too, making more accurate decisions.It will become even more important, helping us make smarter choices across different areas.


Generative Artificial Intelligence(AI)

Generative AI is a fascinating aspect of artificial intelligence that create new content, such as text, images, music, videos, and code, based on patterns learned from existing data. Discriminative Artificial Intelligence(AI) models are already available in Generative Artificial Intelligence(AI) model, so this model,

• Does Not identify difference.

• Does Not solve problems

• Create new information only

 Generative AI deal with unstructured data in 3 domains. 

1. Natural Language Processing (NLP)

NLP is a domain which is interested in textual language. When AI applies to Natural Language or Text, it is called NLP.

2. Computer Vision (CV)

CV is a domain of AI which is interested in Image or video content. It is a branch of AI in which we study methods to solve videos or image problems.

3. Speech Recognition

 It is a domain of AI, which deals with voice such as wavelength, pitch, frequency etc.

Using all above methods, model learns from a wide range of data sources, such as books, articles, images, voices and websites.The model delves deep into the data, grasping grammar rules, contextual relationships, pixel information designs and sound frequencies and stylistic methods. With this knowledge, the generative model can produce fresh content by taking prompts.

Examples of Generative AI in applications:

• Chatbots like ChatCPT and Bard.

• Image Synthesis like DALL-e and Midjourney.

• Music Composition composers like Magenta and Music-LM.

• Video Generation like Runway ML Gen-2. 

Before 2012, domains such as NLP, computer vision (CV), and speech recognition were studied separately. However, with the advent of deep learning, these domains are now integrated, leading to significant advancements in AI capabilities. Both NLP and CV can utilize either discriminative or generative AI techniques, depending on the task at hand.

  • Making a Generative Artificial Intelligence(AI) Model

Generative Artificial Intelligence(AI) requires Generative Artificial Intelligence(AI) Model. To make Generative Artificial Intelligence(AI) Model, in start inputs are given to the computer that is called Seed Input. Seed input is the initial point of generating the  random numbers. 

There are two popular models of  Generative AI

1. Large Language Model (LLM)

This model is related to language so this exists in the NLP domain. These types of models are made up of some specific Neural Network architecture (having billions of parameters or neurons), which is called GPT(Generative Pretrained Transformer). In other words, we can say that LLM is a specific type of transformer called GPT. Working of LLM involves tokenization. 

  • Tokenization

In the working of LLM, tokenization plays a crucial role. It's a process of breaking down the text or prompt into smaller units like phrases, words, or even characters. These smaller units, called tokens, form the vocabulary of the generative AI model. Each GPT model has its own limits of tokens or vocabulary, beyond which the model won’t be able to generate output. The limit of input tokens accessed is called the context window of that LLM

2. Diffusion or Latent Model

Diffusion model operates differently from LLMs. Instead of directly generating content based on learned patterns, it works by diffusing a given input throughout the model's latent space. This diffusion process involves iteratively applying transformations to the input, gradually spreading it out to generate new samples. A simple analogy for this process is spreading ink drop in water. Initially, the ink drop is concentrated, but as time passes, it diffuses and spreads out, creating intricate patterns.



Labels:

Comments

Post a Comment

Popular posts from this blog

Introduction to APIs- OpenAI and Hugging Face

Introduction to APIs - OpenAI and Hugging Face   API: Application Programming Interface is referred to as API. It is a collection of guidelines, conventions, and instruments that facilitates communication between various software programs. APIs make it simpler for developers to integrate features from one program into another by defining how various software components should communicate with one another. They can be utilized to get access to information, features, or services offered by other programs or online services. How do we use the various tools' and websites' APIs?  Through the use of APIs (Application Programming Interfaces) on various websites and tools, developers can incorporate certain features and data from those platforms into their own apps. Here's a condensed explanation along with an illustration: 1. Comprehending APIs:  APIs are similar to collections of guidelines that let various software programs talk to one another. They specify the for...
Post a Comment
Read more

Fundamentals of Machine Learning and Deep Learning

Fundamentals of Machine Learning (ML) and Deep Learning (DL) • Machine Learning Machine Learning is branch of Artificial Intelligence and is popular way to perform AI tasks through various methods. Simply in ML, there is no need to instruct computers, data is given to the system and according to past experience and data , rules are made by computers and apply further to get required results. In this ML domain model of Al are trained on learning how to create general rule for themselves to perform lookalike tasks by taking specific inputs and their desired possible outputs. ML required processed data with human intervention.  Most important things to note that ML algorithms uses neural networks and their hidden layers to process to the final results but these layers  are less i-e 1 or more in ML model, which provides less refined output or results. Other things are that numerous ML algorithms do not respond better on unstructured data as well as can’t handle large amount...
1 comment
Read more