Types of AI
Terms of Use
Except where otherwise indicated, The contents of this slide presentation are available for use under the Creative Commons Attribution 4.0 license.
You are free to adapt and share the work, but you must give appropriate credit, provide a link to the license, and indicate if changes were made.
Sample attribution: [Title of work] by Fred Hutchinson Cancer Center Data Science Lab. CC-BY 4.0
Neural Networks find patterns in hidden layers
Image from Suvanjanprasai, CC BY 4.0, via Wikimedia Commons; Icons from https://www.iconpacks.net/
Input Layer
“Hidden” Layer
Output Layer
Training Data
“0”
CC-by hutchdatascience.org
Deep Learning has many hidden layers
Icons from https://www.iconpacks.net/
Input Layer
“Hidden” Layers
Output Layer
}
“¿Cuáles son algunos recursos donde puedo aprender sobre ciencia de datos?”
“What are some resources where I can learn about data science?”
CC-by hutchdatascience.org
Generative AI
Discriminative AI
Icons from https://www.iconpacks.net/
CC-by hutchdatascience.org
Transformer models help power Generative AI
https://en.wikipedia.org/wiki/Skin_cancer ; Screenshot from Google Scholar.
More than 90% of cases are caused by exposure to ultraviolet radiation from the Sun.[4] This exposure increases the risk of all three main types of skin cancer.[4] Exposure has increased, partly due to a thinner ozone layer.[5][15] Tanning beds are another common source of ultraviolet radiation.[4] For melanomas and basal-cell cancers, exposure during childhood is particularly harmful.[6] For squamous-cell skin cancers, total exposure, irrespective of when it occurs, is more important.[4] Between 20% and 30% of melanomas develop from moles.[6] People with lighter skin are at higher risk[1][16] as are those with poor immune function such as from medications or HIV/AIDS.[5][17] Diagnosis is by biopsy.[3]
!!
“Self Attention”
CC-by hutchdatascience.org
Large Language Models (LLMs) can create text
Icons from https://www.iconpacks.net/
Training Data
CC-by hutchdatascience.org
Large Language Models (LLMs) can create text
Icons from https://www.iconpacks.net/; Text generated by ChatGPT 3.5.
“Describe a nice vacation for winter time.”
Training Data
CC-by hutchdatascience.org
Large Language Models (LLMs) can create text
Icons from https://www.iconpacks.net/; Text generated by ChatGPT 3.5.
“Describe a nice vacation for winter time.”
Training Data
["Describe", "a", "nice", "vacation", "for", "winter", "time", "."]
Encoder
CC-by hutchdatascience.org
Large Language Models (LLMs) can create text
Icons from https://www.iconpacks.net/; Text generated by ChatGPT 3.5.
“Describe a nice vacation for winter time.”
Training Data
["Describe", "a", "nice", "vacation", "for", "winter", "time", "."]
Encoder
Transformer
CC-by hutchdatascience.org
Large Language Models (LLMs) can create text
Icons from https://www.iconpacks.net/; Text generated by ChatGPT 3.5.
“Describe a nice vacation for winter time.”
Training Data
["Describe", "a", "nice", "vacation", "for", "winter", "time", "."]
Encoder
Transformer
Hidden Layers
CC-by hutchdatascience.org
Large Language Models (LLMs) can create text
Icons from https://www.iconpacks.net/; Text generated by ChatGPT 3.5.
“Describe a nice vacation for winter time.”
["Describe", "a", "nice", "vacation", "for", "winter", "time", "."]
Encoder
Transformer
Training Data
Decoder
Hidden Layers
“Imagine a winter retreat in a cozy mountain cabin, surrounded by snow-covered pine trees. You wake up to breathtaking views of glistening snow, spend the day skiing or snowshoeing in the serene landscape, and then relax by the fireplace with a cup of hot cocoa. In the evening, you enjoy a starlit walk, followed by a hearty dinner in the warmth of the cabin. It's a perfect blend of adventure and relaxation, creating cherished memories in the winter wonderland..”
CC-by hutchdatascience.org
Diffusion Models create and remove noise
Apple images generated by StableDiffusionXL. Icons from https://www.iconpacks.net/
Adding Noise
Removing Noise
New Image
Original Image
CC-by hutchdatascience.org
Summary Types of AI
Neural Networks and Deep Learning are both key elements of today’s AI. They function like human brains for advanced pattern recognition.
Transformer architecture is central to many of today’s Large Language Models and allows for rapid processing of context in text.
Diffusion models adjust noise to generate new content, such as images.
CC-by hutchdatascience.org