GPT Under the Hood

Barry S. Stahl

Solution Architect & Developer

@bsstahl@cognitiveinheritance.com

https://CognitiveInheritance.com

Favorite Physicists

1. Harold "Hal" Stahl
2. Carl Sagan
3. Richard Feynman
4. Marie Curie
5. Nikola Tesla
6. Albert Einstein
7. Neil Degrasse Tyson
8. Niels Bohr
9. Galileo Galilei
10. Michael Faraday

Other notables: Stephen Hawking, Edwin Hubble

Favorite Mathematicians

1. Ada Lovelace
2. Alan Turing
3. Johannes Kepler
4. Rene Descartes
5. Isaac Newton
6. Emmy Noether
7. George Boole
8. Blaise Pascal
9. Johann Gauss
10. Grace Hopper

Other notables: Daphne Koller, Grady Booch, Leonardo Fibonacci, Evelyn Berezin, Benoit Mandelbrot

• Armando's Resume: 12.7% "worse"
  • More distant from a match to the job listing
  • A 95 for Jonathon's resume ≈ an 84 for Armando's
• If Armando had 2 additional years of experience
  • His score only increases by ≈ 1 point to 85

• Machine Learning
  • Make predictions based on input
  • Model is tuned via "Training"
• Training
  • Minimize the error in the model

A neural network architecture that has enabled the recent advancements in NLP

• Based on the 2017 Google paper Attention is All You Need by Vaswani et al.
• Key Breakthrough: The Self-Attention Mechanism
  • Focus on most relevant words
  • Add critical context even at a distance
• Enables Parallel Processing
  • Process all words in a sentence simultaneously
  • Unlike traditional RNNs that process data sequentially

Goal: Build intuition for when and why these models may be effectively applied by understanding how they work

• Generative Transformer Models
  • Tokenization
  • Embedding
  • Attention
  • Transformer Blocks
• Use-Case Analysis
  • What they don't work for
  • What these tools do well
  • How we can reduce risk
  • Ethical Considerations

Convert UTF-8 text containing words, word parts, or characters into an equivalent numeric representation

GPT-3 and beyond use the cl100K tokenization model

• Defines 100,256 tokens
  • Word
  • Word part
  • Character combination
  • Character
  • Character part
• Efficiently reprepresent language numerically
  • Minimize token usage
  • More compact than Unicode
  • More flexible than ASCII
  • Large vocabulary
    • All Unicode characters available
  • The most-common combinations

Reference Implementation - AI Demos on GitHub

• Examples from the code
  • "Hello, World! How are you today? 🌍"
    • 9906, 11, 4435, 0, 2650, 527, 499, 3432, 30, 11410, 234, 235
  • "こんにちは、世界！お元気ですか？"
    • 90115, 5486, 3574, 244, 98220, 6447, 33334, 24186, 95221, 38641, 32149, 11571
  • "Hola, mundo! ¿Cómo estás hoy? 🇪🇸"
    • 69112, 11, 29452, 0, 29386, 96997, 1826, 7206, 49841, 30, 11410, 229, 103, 9468, 229, 116

• A point in multi-dimensional space
• Mathematical representation of a word or phrase
• Encode both semantic and contextual information

• Model: text-embedding-ada-002
• Vectors normalized to unit length
• Use 1536 dimensions

Relate vectors based on the angle between them

• Cosine Similarity ranges from -1 to 1, where:

  • +1 indicates that the vectors represent similar semantics & context
  • 0 indicates that the vectors are orthogonal (no similarity)
  • -1 indicates that the vectors have opposing semantics & context

• Cosine Distance is defined as 1 - cosine similarity where:

  • 0 = Synonymous
  • 1 = Orthogonal
  • 2 = Antonymous

Note: For normalized vectors, cosine similarity is the same as the dot-product

Embeddings can be used directly, or as an input to other models

• Direct Usage
  • Measuring Semantic Distance
    • Quantify similarity between pieces of text
    • Useful for tasks like semantic search
  • Clustering for Pattern Discovery
    • Discover groupings in the data
    • Useful when categorizing user comments or other textual characteristics
• Indirect Usage
  • Feature Vectors for Machine Learning
    • Inputs to traditional ML models
  • Input to Transformer Attention Mechanisms
    • Dynamically adjusted by attention blocks
    • Powers text generation tasks

Attention mechanisms consider relationships between words, helping the model emphasize relevant context, even if far away.

• Pre-Attention
  • Does "Ram" represent a sheep, truck or memory?
  • The model has a general understanding of the word without context
• Post-Attention
  • The "Ram" vector shifts towards "Computer Memory"
  • Context has been added

The 3 matrices that are learned in training of the attention mechanism

• Query (Q)
  • Can be thought of as a question about the current token
• Key (K)
  • How well the token "attends to" the Query
• Value (V)
  • How much the embedding should be moved towards each other token

When these 3 matrices are applied to the positionalized input, it results in the model adding appropriate context to each token.

The Feature layers allow the Transformer to understand and generate complex language patterns through deep learning from the structure of the input data

• Each layer acts as a mini neural network
• In the Encoder:
  • Complex Feature Extraction from the input data
• In the Decoder
  • Refines the model's predictions by processing the encoder's output alongside previous decoder outputs

By stacking multiple Feature between Attention layers, GPT models achieve deeper learning capabilities, allowing for more nuanced understanding and generation of text

• Language
  • English (different dialects) vs Spanish
• Historical Ties
  • USA and UK - Political (colonial), economic and cultural ties
• Cultural Similarities
  • USA and UK - shared media, literature, entertainment, and educational exchanges including TV, movies, music, and news
• Political Systems
  • USA and UK - Democracies w/ common roots
• Other differences
  • Economic Ties
  • Geopolitical Alliances
  • Educational Systems
  • Legal Systems

• Clustering
  • Unsupervised machine learning technique
  • Clusters form around centroids (the geometric center)
  • Data points are grouped (clustered) based on similarity
  • Minimize the error (distance from centroid)
• Resume Clustering
  • Identify a group of relevant resumes
  • Not for ranking, just filtering
  • Cluster the resumes, the GPT summaries, or both
  • Still not perfect

• Long Sequence Processing
  • Attention uses O(n²) memory
  • Some models may fragment or chunk the input
  • Be careful with attention over long input
• Local Context Priority
  • Transformers excel at global dependencies
  • Less efficient focusing on local context
• Irrelevant Data
  • Attention may overvalue irrelevant inputs
  • Especially in noisy, unstructured or sparce data
• Resource Constraints
  • Resource constrained devices may not be able to handle Attention
• Low-Resource Domains
  • Transformers require a large training corpus
  • Limited data yields poor attention performance

• Model Creation
  • Copyright violations
  • Energy Consumption
• Model Usage
  • Bias and Fairness
  • Privacy Concerns
  • Integrity
  • Accountability and Transparency

• Copyright Infringement: Models were probably trained on copyrighted materials
  • Consumer Risk: Models removed or nerfed
  • Artist Risk: Theft of copyrighted materials undermines the rights of creators
• Carbon Footprint: Training requires huge resources
  • Energy Usage: Massive compute
  • Water Usage: Cooling data centers
  • Sustainability: Rapid advancement

• Not handling the bias makes bias a feature of your app
  • Understand what fairness means in your context
  • Understand the impact of being wrong
• Take fairness very seriously
  • Prevent all predictable biases
  • Monitor for unpredictable biases

• Incomplete
• Poorly phrased
• Outright wrong

• Massive amounts of private data were likely gathered
  • What are the odds that all of it was caught?
  • Even anonymized data may be de-anonymized
• Take great care in what you send to the Model
  • See Microsoft AI Recall for a recent example

Should this really be done using a stochastic model?

• Why did the model produce the output it did?
• Who is responsible for the answer?
• What is the impact of the answer?
  • What if the answer was wrong?
  • What if the answer was evil?
  • What if humans suffer as a result?

• Use the model to verify it isn't a jailbreak attempt
• Have the model predict the best answer to the query
• Get embedding from predicted answers
• Compare this embedding with a set of pre-defined answer embeddings
• If a close match is found within the pre-defined answers, use it
• If no suitable pre-defined answer is found, request additional information from the user

• When all possible biases have been accounted for
• When all sensitive data has been removed, tokenized or encrypted
• When the stochastic nature of responses has been accounted for
  • A wrong answer is no worse than no answer
  • Outputs have been fully constrained
  • A human is in-the-loop to fix the inevitable errors

• This Presentation - Web | PDF
• AI Code Demos - Code | Docs
• GPT Embeddings: Not Magic - Just Math
• The Depth of GPT Embeddings
• Programmers -- Take Responsibility for Your AI’s Output
• Experiments in Reconstructing Text from Embeddings
• Azure OpenAI Client
• Semantic Kernel
• Carl Sagan - Reasoning on Higher Dimensions (YouTube)

• Discriminative Models
  • Conditional Probability: Predict label Y based on input X
  • Identify the boundaries between data groups
  • Great for classification and tokenization
  • Examples: BERT, Random Forest
• Generative Models
  • Joint Probability: Predict when label Y occurs along with input X
  • Predict new data instances
  • Great for predicting text, creating audio & video
  • Examples: GPT-4o, PixelRNN/PixelCNN

SoftMax(z) = e^(z_i) / Σ(e^(z_j))

• Converts a vector into a probability distribution
• Output probabilities range from 0 to 1
• Output probabilities sum up to 1
• Where
  • z is the input vector
  • i is the class being predicted
  • j runs over all classes

f(x) = max(0, x)

• ReLU (Rectified Linear Unit)
  • Outputs the input directly if it is positive
  • Otherwise, it outputs zero
• Common activation function used in neural networks
  • Very fast and efficient
  • Helps avoid the vanishing gradient problem
  • Training often converges faster vs sigmoid or tanh

• Store/retrieve high-dimensional vectors
• Values are retrieved using similarity searches
• Leverage data-structures such as K-D Trees
• Examples
  • Azure AI Search
  • Redis
  • Qdrant
  • Pinecone
  • Chroma