Connectionism Flashcards
Understanding the limitations of the computationalist paradigm seen from the connectionist paradigm 2) Being able to reflect on the values of thought experiments 3) Understanding how neural networks work on a broad level 4) Understanding how semantic content may emerge from the network, i.e. how symbols may get grounded
Connectionism: A paradigm shift, from Computationalism
In computationalism there was a belief that the brain was like a computer and that a range of behavior was made possible by internal computation.
In connectionism, a small shift happened, instead of internal computation it was believed that behavior was possible by neural networks, and that the human mind’s processes lay in the neural networks.
What is the luminous room?
Just explain the thought experiment, not how the argument works.
David Cole’s “Luminous Room” is a counter-thought experiment intended to challenge Searle’s conclusion.
Imagine a person in a dark room, shaking a magnet up and down.
According to the laws of electromagnetism, this motion creates an electromagnetic wave.
If you were unaware of these laws, you wouldn’t perceive the creation of the wave just by observing the person’s actions.
Explain the luminous room argument. (Not explain what the argument is in itself)
David Cole’s “Luminous Room” is a counter-thought experiment intended to challenge Searle’s conclusion.
Cole’s argument hinges on the concept of emergent properties. Just as electromagnetic waves can emerge from physical motions, consciousness or understanding might emerge from complex symbol manipulations within a computer.
What is the three axioms and the conclusion in the Luminous room?
Axiom 1: Electricity and magnetism are forces
Axiom 2: The essential property of light is luminance
Axiom 3: Forces by themselves are neither constituitive of nor sufficient for luminance.
Conclusion: Electricity and magnatisme are netiher constituitive of nor sufficient for light.
Name the three arguments against the luminous room. Only headline
- Axiom 3 is question-begging
- Our intuitions doesn’t constrain nature
- The question is empirical
What was Searle’s rebuttal to the luminous room?
He believes the analogy fails because: the electromagnetic radiation is a causal story; on the contrary, formal symbols have no causal power.
* Searle claims that electromagnetic radiation results from a well-understood causal process: the movement of a magnet generates electromagnetic waves according to the laws of physics.
* This is a physical, causal phenomenon with measurable, observable effects.
What is the “Axiom 3 - question begging” argument against the luminous room?
Question begging: An attempt to prove something is true while simultaneously taking that same thing for granted.
* Example: It is time to go to bed, why?, because it is your bedtime.
On the background of this, they believed that the third axiom made is almost the same as the conclusion, and the axiom and the conclusion cannot be the same thing, as it then has no solid foundation to build its conclusion on.
What is the “Our intuitions doesn’t constrain nature” argument against the luminous room?
Meaning; The way we humans think about how something works, doesn’t mean it is the way the natural world actually operates. So the natural world has its own laws and principles, regardless of how humans thinks it should be.
What is the “question is empirical” argument against the luminous room?
As part of why there isn’t light, is because the human cannot move that fast to create it in the given thought experiment, but also it is an empirical question. Cause even though the human eye might not be able to see if there in fact is a little bit of light, the human eye might not be able to see it, even though there might in a more objective way (different wavelength) of light invisible to the human eye might in fact be there.
The Context problem - Human cognition?
**Flexibility and Adaptability: **Humans can effortlessly switch and integrate contexts, utilizing a vast amount of implicit and explicit knowledge.
Intuitive Understanding: Human cognition leverages both logical reasoning and intuitive, experience-based understanding to manage context dynamically.
The Contex problem - Classical AI
Explicit Context Management: Classical AI systems manage context through predefined structures and rules, making them robust in well-defined environments but less flexible in dynamic, real-world situations.
Limitations in Learning: Classical AI struggles with learning and adapting to new contexts without manual intervention, limiting its ability to generalize from experience as humans do.
Whatis the connectionist networks? Reverse engineering the brain
Brains are parallel machines
* While machines are often serial.
Neurons, its processing units are analogue
* It can add a number between 0 or 1
Projects are both feed-forward and recurrent
The blue lines connecting the different layers are weights. So, the top blue might have a stronger connection to the top yellow than some of the other yellows, and are more likely to make a connection to the top than the others.
What are the afvanteses of parallelism in the connectionist network?
Speed
Redundancy - Fault tolerance
* The system doesn’t fail as easily
Flexible storage
What did NETtalk do?
NETtalk - One of the first machines which could recognize an input text (a letter) and then connect it to the phonemic sound.
What is gradient descent used for?
Gradient descent is used to train machines to learn to give the correct output.
through reinfored learning.
