LECTURE 1
Main goals:
● Define generative AI
● Explain how generative AI works
● Describe generative AI model types
● Describe generative AI applications
Define generative AI
Generative AI is a type of AI technology that can produce various type of content including:
● Text
● Imagery
● Audio
● Synthetic data
Differences of AI and machine learning:
● AI is a discipline that has to do with the theory and methods to build machines that
think and act like humans.
● ML is a subfield that gives the machine the ability to learn without using explicit
programming.
○ Supervised:
■ Labeled data: tags such as name, Data set 49.
○ Unsupervised
■ Unlabeled data: a data that comes with no tag
● Deep learning
LECTURE 1
Machine Learning:
Supervised Learning
Unsupervised Learning
Reinforcement Learning
Deep learning: Generative Ai it is a subset of deep learning which means it uses artificial
neural networks can process label and unlabeled data, using supervised, semi-supervised
and unsupervised lessons.
Machine learning models can be divided to two types:
1. Generative model: generates new data instances based a learned probability
distribution of existing data. Generative models generate new contents
2. Discriminative model: used to classify or predict labels for data points. They are
trained on the data set of labeled data points and they learn the relationship
between the features of data points and the labels. Once a discriminative model is
trained it can be used to predict the label for new data points.
Examples
LECTURE 1
What is Generative AI and why not?
Specifically:
LECTURE 1
Definition: Generative AI is a type of Artificial Intelligence that creates new content based
on what it has learned from existing content.
Generative language models learn about patterns in language through training data, then
given text, they predict what comes next.
Generative image models produce new images using techniques like diffusion, then gives an
prompt or related imagery they transform random noise into images from photos.
Transformer model
Hallucinations are words or phrases that are generated by the
model that are often nonsensical or grammatically incorrect.
This can be caused by a lot of factors such as:
When the model is not trained on enough
data
When the model is trained on noisy or
dirty data.
When the model is not given enough
context or is not given enough constraints.
What is the outcome?
LECTURE 1
Hallucinations make the output text difficult to understand and make
the model more likely to generate incorrect or misleading
information.
Prompt: is a short piece of text that is given to a large
language model (llm) as input and it can be used to
control the output of the model in a variety of ways.
Model types available:
Text → Text: they take a natural language input and produce a text output, these
models are trained to learn the mapping between a pair of text i.e. translating from
one language to another.
Text → Image: are trained on a large set of images each captioned with a short text
description. Diffusion is one method used to achieve this.
Text → Video: aim to generate a video representation from text input, the input text
can be anything for a single sentence to a full script and the output is a video that
corresponds to the input text.
Text → 3D models: generate three-dimensional objects that correspond to a user’s
text description for use in games or other 3D worlds.
Text → Task: are trained to perform a defined tasked or action based on text input
this task can be a wide range of actions such as answering a question, performing a
search, making a prediction or taking some sort of action i.e. navigating a web user
interface or making changes to a doc through a graphical user interface.
Foundational Model:
A large AI model pre-trained on a vast quantity of data designed to be adapted or fine tuned
to a wide range of downstream tasks such as sentiment analysis, image captioning and
object recognition. They can revolutionize finance, health care and customer service.
LECTURE 1
Language Foundational Model:
PaLM API for chat
PaLM API for text
BERT
Vision Foundational Model
Embedding extractor
BLIP image captioning
Stable Diffusion v1-5
BLIP VQA
CLIP
OWL — ViT
ViT GPT2
LECTURE 1
Vertex AI studio
Quickly explore and customize.
Developers create and deploy
o Library of pre-trained models.
o Tool for fine-turning models.
o Tool for deploying models for production.
o Community forum for developers to share ideas and collaborate.
Vertex AI
Chatbots
Digital assistants
Custom search engines
Knowledge bases
Training applications
PaLM API
model training tool — train ml models on their data using different algorithms
model deployment tool — deploy ml models to productions with a number of
different deployment options
model monitoring tool — monitor performance of ml models in production using a
dashboard and a number of different metrics.
