Generative AI learning path by Google

Introduction to Generative Artificial Intelligence

Introduction

• Al, Artificial Intelligence is the theory and development of computer systems able to perform tasks usually requiring human intelligence.

• ML, Machine Learning, allows computers to learn without explicit programming.

• DL, Deep learning uses Artificial Neural Networks (ANNs) - allowing them to process more complex patterns than traditional machine learning.

• Generative AI and LLMs are a subset of Deep Learning.

Deep Learning Methods

• There are two types of DL methods:



• Discriminative:
• Used to classify or predict
• They are typically trained on a dataset of labeled data.
• Learns the relationship between the features of the data points and the labels
• Generative
• Generates new data that is similar to data it was trained on
• Understands the distribution of data and how likely a given example is
• Predict the next word in a sequence.

 

What is Generative Al?

• Generative Al is a type of Artificial Intelligence that creates new content based on what it has learned from existing content.

• Learning from existing content is called training and results in creating a statistical model.

• When given a prompt, GenAl uses this statistical model to predict an expected response and generate new content.

• Generative language models learn about patterns in language through training data. Then, given some text, they predict what comes next.

Hallucinations

• Hallucinations are words or phrases generated by the model that are often nonsensical or grammatically incorrect.

• Challenges:
• The model is not trained on enough data.
• The model is trained on noisy or dirty data.
• The model is not given enough context.
• The model is not given enough constraints.

Generative AI model types

1. Text to text
1. Generation
2. Classification
3. Summarization
4. Translation
5. Extraction
6. Clustering

2. Text to image
1. Image generation
2. Image editing/expanding

3. Text to Video/3D
1. Video generation
2. Video editing
3. Game assets

4. Text to Task:
1. Text-to-task models are trained to perform a specific task or action based on text input. This task can be a wide range of activities, such as answering a question, performing a search, making a prediction, or taking some action. For example, a text-to-task model could be trained to navigate web UI or change a doc through the GUI.
2. Types:
1. Software Agents
2. Virtual Assistant and
3. Automation

Reading Materials

Introduction to LLMs

• Large, general-purpose language models can be pre-trained and fine-tuned for specific purposes.

Attributes of LLMs:

• Large
• Large training dataset
• A large number of parameters

• General purpose
• The commonality of human languages
• Resource restriction

• Pre-trained and fine-tuned

Benefits

1. A single model can be used for different tasks.

2. The fine-tuning process requires minimal field data.

3. The performance is continuously growing with more data and parameters.

LLM Types

1. Generic (or Raw) Language Models | These predict the following word (technically token) based on the language in the training data.

2. Instruction Tuned | Trained to predict response to the instructions given in the input.

3. Dialog Tuned | Trained to have a dialog by predicting the subsequent response.

Tuning

• Tuning: The process of adapting a model to a new domain or set of custom use cases by training the model on new data. For example, we may collect training data and "tune" the LLM specifically for the legal or medical domain.

• Fine-tuning: Bring your dataset and retrain the model by tuning every weight in the LLM. This requires a big training job (like tremendous) and hosting your own fine-tuned model.

Reading Materials

Introduction to Responsible AI

8 AI Principles by Google

1. AI should be socially beneficial.

2. AI should avoid creating or reinforcing unfair bias.

3. AI should be built and tested for safety.

4. AI should be accountable to people.

5. AI should incorporate privacy design principles.

6. AI should uphold high standards of scientific excellence.

7. AI should be made available for uses that accord with these principles.

8. In addition to these seven principles, there are specific AI applications we will not pursue.
1. We will not design or deploy AI in these four application areas: Technologies that cause or are likely to cause overall harm.
2. Weapons or other technologies whose principal purpose or implementation is to cause or directly facilitate injury to people.
3. Technologies that gather or use the information for surveillance violate internationally accepted norms.
4. And technologies whose purpose contravenes widely accepted international law and human rights principles.

Introduction to Image Generation

Encoder-Decoder Architecture

Attention Mechanism

• The attention mechanism is a technique that allows the neural network to focus on specific parts of an input sequence.

• Check the following for more Attention.

Transformer Models and BERT model

• Backpropagation at the self-attention layer calculates query, Key, and Value vectors.

• BERT: Bidirectional Encoder Representations from Transformers

• BERT is trained on two different tasks.

• Each word is represented in token embeddings.

• How does Bert distinguish the input in a given pair? By using segment embeddings.

• The order of the input sequence is incorporated into the position embeddings.

Create Image Captioning Models

Introduction to Generative AI Studio

Further, Read