Week 1 Flashcards
prompt engineering
getting the most out of the AI model through clear communication
What makes a strong prompt engineer?
- clear communtication
- willingness to iterate
To use AI successfully, you need to be…
smart, focused, think critically, and proficient in reading and writing
Most underrated prompt
the ability to identify and articulate your implicit knowledge is the true differntiator between average and excellent prompt engineers
Arguments for calculators not being a form of AI
- Following fixed rules, giving fixed answers (intelligence is about giving open-ended answers based on the context – e.g. pointing)
- Passively waiting for human inputs (intelligence is about responding to the environment autonomously and adapably – robot dog responding to being kicked)
- No learning, every rule is manually defined by humans
Communication as making a point
Some parts of communications are so innate for us that they’re difficult to model
e.g. difficult for AI to intepret the meaning of a point (e.g. a man pointing at a shot glass), however its ability to do so has improved over time
Two dimensions of AI
- internal thought processing and reasoning and external behavior
- fidelity to human performance and an abstract, formal definition of rationality
Turing test
a test for intelligence in a computer, requiring that a human should be unable to distinguish between the answers from a human and from AI
- passes the test if a human interrogator, after posing some written questions, cannot tell whether the written response come from a person or from a computer (e.g. in a text exchange)
What does and doesn’t the turing test focus on?
-does focus on the outcome
-doesn’t focus on how the machine works
Limitations of turing test
Can’t execute tasks in the physical world because executing tasks is more difficult than writing and creating art
Current results of turing test
People cannot distinguish GPT-4 from a human in a turing test
How does AI use cognitive modeling?
Uses cognitive modeling as reverse engineering in the human mind
reverse engineering
sees the product and then replicates it
*involves taking the product apart, understanding how it works, making improvements, the rebuilding it with improvements
false belief task
-a psychological test used to assess an individual’s theory of mind—the ability to understand that others can hold beliefs different from one’s own
-it evaluates whether a person can recognize that someone else might believe something that is untrue
(a “false belief”) based on incomplete or misleading
information, even if the person taking the test knows the
actual truth
e.g. box and basket example, ramen in microwave example
paternalistic helping
We can predict other behavior based on THEIR knowledge
AI Rational Approach
AI comes from statistics and data
What is the next generation of AI
Embodied AI: Perceiving and interacting with the physical world
ImageNet Classification
database used for visual object indentification
Geoffrey Hinton won Nobel Prize in physics
What is something AI currently struggles with?
learning how to count from sensory inputs
e.g. creating five fingers on a hand
What is AI human and acting overlap
-turing test
-embodied AI
What is AI human and thinking overlap
-cognitive modeling
-false belief task