Introduction to Generative AI

Define gen Ai : A type of generative technology which generates content like text image audio

Ai > mL > DL > gen Ai/descriptive Ai

AI : Ai is descipile of compute science which deals with creation of intelligent agent and a system that can reason, learn, act autonomously.

mL :

Machine Learning is subfield of Ai, and is a program or system that trains a model from the input data, which model gives machine a ability to predict on the never seen before data without explicit programming.

Supervised models : uses Labeled data

tried to minimise that error comaparing the expected outputs

Unsupervised models : uses unlabled data ( patterns are discoverd naturally )

Deep Leaning

Deep learning comes under the machine learning where it used neural network to recognizsed/process the more complex patters than the machine learning.

Neural network can use both labeled and unlabeled data ( semi-supervised learning ), where small label data helps to learn the basic concepts or to form those foundation patters and unlabled data helps to generalise the neural network for new/unseen examples.

Explain gen ai works

Gen Ai

Generative Ai is subset of deep learning, Deep learning/ machine learning models can be divided into the discriminative model or Generative model.

Discriminative models are typically trained on labeled data to form a relation between the imput data and the output.

Generative Model generates the content based on the learned probability distribution of existing model.

what is gen ai or what is not?

Gen ai outputs > natural language (text) / image/ audio

Not Gen AI output -> number/discrete/class/probability

Transformers

Transformer model -> consists of encoder and decoder.

issues by transformer -> Hallusinations.

Hallucinations are sentences or phrases which are non-scensical or gramatically not correct.which may lead to bad output and hard to understand or mislead the information. this could cause due to lack of data or noisy data while training, or context.

Prompt

prompt is small text that is given to the model as input.

DIfferent types of model

text to text

text to image -> which is trained on the large set of images with labeled text. ( Diffusion model )

text to 3d or video

Text to Task ->

Foundation model

A a model which is trained on the vast quantity of data which is designed to be adapted or fined tuned to a wide range of downstream (more specific ) tasks

get ai applications