Artificial Intelligence Flashcards
Minds and Computers
Both use binary electrical signals to process inputs and respond with some type of output
(True) Artificial Intelligence
capable of consciousness, learning on their own, able to make decisions on their own will, would be unpredictable (like a human is)
Virtual intelligence
Not truly conscious - governed by a set of programs… everything called AI is really just virtual intelligence
The Turing Test
Computers will have achieved true artificial intelligence if a human being could not tell if he/she was talking to a computer or another human being
Turing Test Accuracy
Pretty easy to fool a human being
Brains are better than computers
Digital vs Analog
- Computers operate entirely digitally (0 or 1)
- Brains are a combination of digital and analog
Brains are a combination of digital and analog
Each neuron either fires or doesn’t (digital)
But… signals of thousands of individual neurons create continuous “waves” of electrical and magnetic activity (analog)
Serial
Computers calculate one thing at a time (serial)
- Switches back and forth between tasks to multi-task
- *Parallel processing in computers involves breaking a single problem into multiple parts that can be solved simultaneously
- But… it is still only solving one problem at a time
Parallel
Brain processes many different things simultaneously
- Your body runs totally independent of your conscious mind, but its still your brain controlling it all
- Brain is leaky - how you process one thing effects the next thing
- Brain is parallel impact of one thing influences later calculations
Deterministic vs. Non-deterministic
- Computer output determined by its programming
- If you know a computers inputs, you know its outputs
- “random” numbers from a computer are an algorithm too
Brains are non-deterministic
Same things causes different responses at different times
- Hard to explain human behavior as following a set of rules*
- Unless you have a deterministic view of free-will
- Dynamic Self-Construction vs Static Construction
- A computer cannot rebuild, rewire, or redesign itself
- The physical form of the computer can not change based on learning or experience
Speed and Repetition
- THe fastest neuron can fire is 1000 a second
- COmputers can do 100 quadrillion calculations per second
Memory Storage and Capacity
- Can be connected to store unimaginable amounts of data
- Computers don’t forget
Dimensions of calculation
- Human brains have difficulty thinking in more than 3 dimensions
- Computers can do calculations involving thousands of dimensions
- Computers can find patterns in large data sets that the human mind can no
Artifical Neural Networks (ANNs)
- Artifical Neural Networks (ANNs)
- Attempt to create “neuron-like” networks
- Can be hardware or software based
- Capable of learning
ANN Learning
- they learn from experience
- Each “neuron” in the network is just a numeric calculation
- But they are connected. Changes in one affects others
- Different layers to stimulate how brain processed information
Artificial Neural Networks Operations
Input Layer
- The input layer translates the stimulus into a collection of numbers
- Pixels in the image become a set of numbers
“Hidden Layers”
- Hidden units keep track of how “neurons” were activated
- Patterns in the data are found by changing values based on feedback
Output layer
- How the network responds
Back-propogation
- Network adjusts based on feedback about error signal
- Changes in values in the hidden layers
ANN use reinforcement learning
- The network’s “behavior” is shaped by feedback it gets
- Correct outputs are “rewarded,” incorrect outputs are “punished” (error signal)
- The process repeats itself until the error signal is minimized
- When the error is minimized, the network has properly learned the category or concept
Generative Adversarial Networks (GANs)
- One ANN attempts to generate a stimulus (the generator)
- Another ANN is used to categorize whether the generator was successful
- For instance, a network trained to recognize cat photos could recognize whether a different network created a decent cat photo
Consciousness as emergence
- Where is consciousness in the brain?
- There is no leader or director of conscious awareness
- No single neuron could be considered conscious on its own
- Emergence
- When new abilities are created by interaction of simpler units which themselves don’t have the ability
- Billions of Neurons firing creates consciousness
- Billions of chemical interactions create life
