Trends Final Flashcards
What is AI and what are the 3 types
Artificial Intelligence
Generative
Predictive
Stochastic Parrot
Disruptive Technology
is an innovation that significantly alters how consumers, industries, or businesses operate. It supersedes an older established process, product, or habit with recognizably superior attributes.
ML
Machine Learning
uses data and algorithms to mimic
human learning
statistical methods to train
algorithms to classify or predict and
even provide insights into data
mining projects
Generative AI
Generative AI focuses on understanding patterns and structure in data and using that to create new data that looks like it. This includes writing blocks of text, lines of code or creating photorealistic images
Predictive AI
Is mainly on classification, learning the difference between “things”
This is what’s used in recommendation engines like those used by Netflix or Amazon to distinguish between things you might want to watch or buy and things you’re unlikely to be interested in
3 major features of AI
Extremely high computing speed
Large volume dataset processing
ability to self learn
3 levels of AI
ANI
AGI
ASI
ANI
Narrow intelligence
is specialized
to the function for which it has been developed
Specific
AGI
General intelligence
Is generally referred to as
‘human-level AI’, because it describes the capacity of a computer that is as smart as
a human, a point often referred to as ‘Singularity’
ASI
Super intelligence
ASI is the point at which computers
possess an intellectual capacity far greater than that of human beings with the capacity for social skills and general knowledge that would increase exponentially over time.
3 examples of AI
Chat gpt
Luna
soudstorm
LLM
Large language models
artificial intelligence program designed to understand, generate, and work with human language on a large scale
NLP
Natural language processing
involves enabling computers to
understand, interpret, and respond to
human language in a way that is both
meaningful and useful.
Application programming interface
protocols that allow different software programs to communicate with each other/AI
Deep Learning
subset of ML
uses neural networks to analyze and
learn from data.
Randomness vs learning and data volume
“Monkey in the theorem”
based on generating text randomly, without any understanding
output mostly random gibberish with the occasional coherent sentence
no training
Ai
learning from the previous inputs.
Understanding vs mimicking
“Monkey in the theorem”
do not understand the meaning of the text it is generating
AI
AI and NLP models do not truly understand the text they generate
3 strategies for facial recognition
feature based
appearance based
knowledge based
Feature based facial recognition
identify fiducial points: specific facial landmarks
measures relationships and distances between facial features
generally more robust to variations in lighting and facial expressions since it relies on stable and distinctive facial landmarks
Appearance based facial recognition
Appearance-based methods use holistic information from the face.
Instead of focusing on individual features, these techniques analyze the entire facial image as a whole
can be more effective in capturing more detailed and subtle facial characteristics
generally easier to implement since they don’t require the detection of specific facial landmarks
knowledge based facial recognition
designed to identify suspicious or abnormal behaviors based on predefined rules or knowledge about facial expressions and behaviors.
relies on a database of facial expressions/ behaviors corresponding with specific intentions/emotions
3 benefits of facial recognition
Enhanced security and safety
Efficiency in policing
finding missing persons
3 disadvantages to facial recognition
Privacy concerns
Potential for abuse
Bias and inaccuracy
AFIS
Automated fingerprint identification system
how many records does the AFIS database contain
4 million
What does AFIS do
It attempts to locate and measure the reliable and persistent features within fingerprint and palm print images
It then notes their location X-Y coordinates
And their theta (θ) vectors – showing the orientation of those features.
i.e. ridge endings, bifurcating ridges, large ridge dots
Can also use minutiae, sweat pores and edge features
Can fingerprints be the same
No two complete fingerprints have been found to be the same – Not
even from the same person.
4 problems AI overcomes in the AFIS database
growth
injury
deviation
quality of record
CPAP
Capture, process, analyze, present