1 / 11

Minds and Machines

Minds and Machines. Summer 2011 Wednesday, 8/3. Biological Approaches to Understanding the Mind. Connectionism is not the only approach to understanding the mind that draws on Biology, or the actual organization of the brain.

maura
Download Presentation

Minds and Machines

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Minds and Machines Summer 2011 Wednesday, 8/3

  2. Biological Approaches to Understanding the Mind • Connectionism is not the only approach to understanding the mind that draws on Biology, or the actual organization of the brain. • In his chapter, Clark shows how our knowledge of the brain has led us to move away from an influential, computational paradigm for the study of vision. • This is significant because: if studying the brain can help us understand one mental process, then it is reasonable to expect it to illuminate many others.

  3. Marr’s Levels of Analysis Level 1 (task): Characterize the task being performed (e.g. identifying 3-d objects via 2-d inputs) Level 2 (algorithm): Describe a scheme for representing the inputs and outputs and a sequence of mechanical steps that would carry out the task. Level 3 (implementation): Determine how to build a device that would run through the sequence of steps.

  4. Marr’s Levels of Analysis • Until the 80s, many cognitive scientists took this framework to warrant ignoring or downplaying the importance of understanding the brain. • Nowdays, most cognitive scientists agree that discovering good computational models of cognition should be informed by neuroscience. • Still, there is agreement that there was something deeply right in Marr’s approach. We won’t understand the mind just by looking at what goes on in the brain. We need a general understanding of what the brain does, or what computational strategies it implements.

  5. Evolution vs. Human Engineering • The computational processes that we are tempted to posit are likely to diverge from the evolved computational processes our brains actually carry out. • Evolution is both constrained and liberated in ways human engineers are not. • It is constrained because it builds its solution incrementally by a process of “tinkering”. • It is liberated because it is able to discover “messy” and unobvious solutions that would baffle human engineers.

  6. Evolution vs. Human Engineering • For example, consider the problem of controlling finger motion in monkeys. • A human engineer may come up with a neat solution where dedicated groups of neurons individually control each finger. So you would expect more neurons to fire when a whole-hand movement (e.g. a grasping movement) takes place. • It turns out that exactly the opposite happens: moving individual fingers requires more activity (to inhibit aspects of whole-hand movement). When you think about it from an evolutionary perspective, it makes a lot of sense! • This is also a good illustration of how evolution solves problems by tinkering (in this case with coordinated whole-hand commands).

  7. Vision: the traditional view • The function of vision is to produce detailed inner representations of the 3-d visual scene on the basis of (impoverished) 2-d retinal images. • Such representations are then given as inputs to reasoning and planning centers, whose job is to determine a course of action and then send commands to the motor areas to carry it out.

  8. Chipping away at the traditional view • Psychology and neuroscience gives us reasons to think that the visual-system retrieves information as and when the information is needed for some specific problem solving purpose. • Our sense that we enjoy rich and detailed visual experiences, that every part of the visual scene enters our awareness, is a grand illusion! (Compare: the illusion that everything is there in a store that uses a computer ordering system)

  9. Chipping away at the traditional view • We can design machines that can perform complex tasks that require sophisticated environmental sensitivity, without requiring rich representations of the environment or advance planning. • E.g. Herbert: walks around randomly, can avoid basic obstacles, detects outlines of cans using laser, can put himself in a standard position in front of a can and picks it up. • This requires no complex inner model of the environment. Sometimes: “the world is its own best model”.

  10. Chipping away at the traditional view • Actions may play a role in the computation process that leads to visual outputs. Sometimes action guides vision rather the other way around! • For example, the process of distinguishing figure from ground uses information obtained from head movement during eye fixation. • The process involved in depth perception uses cues obtained by the observer’s motion towards the object.

  11. Chipping away at the traditional view • Neural representations of events in the world may themselves already be recipes for action rather than passive data-structures that are given as inputs to reasoning processes. • For example, there are neurons in the monkey’s ventral pre-motor cortex (called mirror neurons) that are active both when the monkey observes a specific action and when the monkey performs the same kind of action. The perceived action is stored in terms of an action code, not a perceptual code. • Such representations may describe the world by depicting it in terms of possible actions (e.g. visual experience of artifacts, Phenomenology).

More Related