Artificial Intelligence (AI)
leveraging AGI to solve specific objectives and problems
Artificial General Intelligence (AGI)
A 1 to 1 from machine to human use cases
Text-Related Artificial Intelligence Purpose
used to extract and understand text so it can generate
Text-Related ML Process
Tokenization of words → Padding →Embedding
Tokenization
converting words into numbers
Padding
length of sentences can vary so we pad shorter sentences with 0’s for vector-computational efficiency
Embedding
the ID from Tokenization becomes a vector to be compared for similarity
Visual of Tokenization → Padding → Embedding
“Hello World” → [101,2023,102] → [101,2023,102,0] → [(.1,.3.-2),(2,3,4),(…)]
Recurrent Neural Networks
processes data sequentially and stores hidden states
Long Short Term-Memory
processes data sequentially and can retain the context better w/ gates
Transformers
processes data in parallel. Uses concept of self attention to better understand
Variational Auto-encoders (VAE)
compresses data into a latent space and reconstructs it while learning a probabilistic representation
Waveform
neural networks designed to generate or process raw audio directly
Siamese
two twin networks w/ shared weights that learn to measure similarity between pairs of inputs by comparison w/ each other
Convolutional Neural Networks
detects patters in images and learn hierarchical representations
YOLO
process the image and detects objects within the image
Generative Adversarial Networks
generates real looking images
Machine Learning
algorithms learned from past data and predict outcome on new data
Machine Learning Types
Supervised, Unsupervised, Reinforcement Training
Supervised Machine Learning
extracting rules from existing data
Unsupervised Machine Learning
extract trends to get insights
Reinforcement Training
use a “credits” system to determine good/bad actions
Deep Learning
uses complex data w/ neural networks to predict outcomes and generate data
