Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
46 Cards in this Set
- Front
- Back
Provide both definitions of Artificial Intelligence. |
The field that attempts to design an intelligent machine. An intelligent machine (an AI) |
|
Why do we think we can create AI? |
Functionalism: What makes something mind is not that it's made of, but how it functions. Mental states are physical states. Anything that can implement those physical states is a mind (nature of physical system doesn't matter). |
|
What is a computational task? |
One that follows a well-defined process/set of steps. |
|
What is an Algorithm? |
Sequence of instructions, effective procedure. |
|
What is the claim of AI? |
Cognitive tasks can be reduced to computational tasks. |
|
What is a node in an ANN? |
Units in network, they have activation values (numbers associated with that node). Also have thresholds, can fire if input exceeds threshold. |
|
What is a link in an ANN? |
Connection between nodes; have weights (number associated with that link) |
|
What values can a weight have in an ANN? |
between 1 and -1, closer to 1 or -1 = 'heavier' weight. |
|
What is the basis function for ANNs? |
Equation describing how inputs to node are calculated. Sj = (SUM)AiWji - Each connected node's activation value (Ai) multiplied by connection weight of link between the two nodes (Wji) |
|
What is an Activation function in an ANN? |
f(net): An equation describing how output affected by input. |
|
What is Backpropagation? |
Method of training ANN by gradually changing weights on links in response to feedback. |
|
What is the Backpropagation process? Step by step. |
Set all weights, Set target output for output nodes, Run ANN get actual output nodes, Subtract actual from target output(Error signal = target output - actual output), send error signal back through network. |
|
What is a Dynamical System? |
System that evolves over time in law-governed way. |
|
Define a State space. |
Geometrical space with as many dimensions as there are variables. |
|
How is cognition treated through the DST? |
Treated as outcome of set of interacting variables. Order emerges from chaos. Cognition = trajectory through state space. |
|
ANNs vs DST |
Complementary ways of modeling mind. |
|
Classical AI vs ANNs & DST |
Emphasis on setting initial conditions, letting system learn/evolve. Physical and social worlds have huge impact on cognition. |
|
How to get physical and social worlds into AI? |
Give Artificial mind Artificial body! i.e. Robotics |
|
What are the 3 building blocks of classical robotics? |
Sense, plan, act. |
|
How do robots sense? |
Sensors: Things that detect change in environment. Knowledge base: Specific info programmer gives system. |
|
How do robots act? |
Send control signal to actuators (parts of machine that move/ control a system) ex. Move wheels. |
|
How do robots plan? |
Use info from sense, generate task to execute. ex. go down. |
|
Explain sense, plan, act |
Compare current state to knowledge base/sensors, generate plan, execute plan, repeat. |
|
What is another approach to classical robotics? |
Subsumption architecture, get rid of planning, just sense and act. |
|
What is the uncanny valley? |
When graphing affinity & resemblance to human, don't see continuous increase, see large dip. |
|
How are ANNs, DST and robotics connected? |
Many robots use ANNs, DST. DST is current best proposal for embodied, distributed robotics and maybe human cognition? |
|
What is problem solving? |
Navigation through problem space to find solution. Usually: have initial, know/guess final state, guess intermediate states. |
|
What is a problem space? |
Mental representation of problem, including initial, final and possible intermediate states. |
|
Why can computers solve some problems much faster than humans? |
Mind is not a perfect machine. |
|
Why are some human problems hard for computers to solve? |
Many human problems aren't well-specified, don't come w/ finite set of instructions. |
|
Do external factors affect human problem solving? |
Yes. Messy vs. tidy work area. Chronotypes |
|
What are two different types of problems? |
Insight: solved suddenly with an AHA! moment. Analytic, Use logic, trial & error, search through problem space. |
|
What is analogical reasoning? |
Using knowledge about one situation to reason about another situation. |
|
What is a relational pattern? |
Sameness acress two situations. |
|
What is mapping? |
Transferring info from one domain of experience to another. |
|
What are Target and Source when it comes to mapping? |
Source: Usually well understood, concrete, familiar, the thing you're using to understand the other thing. Target: Usually NOT well understood, abstract, the thing you're trying to understand. |
|
How are analogies written? How are they mapped? |
Target -> source, source -> target |
|
What is Inferencing? |
Making new conclusions based on mappings from source to target. |
|
What is Evaluating? |
Deciding which info Not to map (which inferences not to draw) |
|
Summary of Analogy |
Major method of problem solving for humans. We're better at it than computers. Sameness (relational patterns) in mind, not world. |
|
What is Design? |
Creating plan for the construction of an object or a system. |
|
Where do design failures arise from? |
Limited capacity of memory – Performance affected by external factors (Tme of day, stereotypes, environment) – Theory of mind (user & designer’s mental states don’t match) – Distributed cogniTon (one person’s cogniTon clashes with another’s). – Objects designed to support epistemic actions make thinking harder... – Embodied understanding of objects not considered (bad handles) |
|
What are the 4 principles of good design? |
Have visible cues to operation of object (no false cues); Take advantage of affordances (knowledge about how an object can be used, interacted with, must treat cognition as embodied); If extra stuff (diagrams, instructions) needed, design has failed; Good design induces flow. |
|
What is flow? |
State of consciousness, characterized by intrinsic enjoyment, sustained attention, loss of self-consciousness, Presence: sense that you have left real world and are present in virtual world. Can apple to simple object, more often used for activity. |
|
What are some flow-inducing activities? |
The internet, games, writing a paper, debugging code, having a conversation, playing a game |
|
What are some properties that induce flow? |
Clear goal, challenges that match your skills, control over the task, immediate and efficient feedback. Bad design bad for flow. |