Scientific Lives: Desires, Diversions, and Styles

Scientific Lives: Desires, Diversions, and Styles
paly

This text explores the varied desires and diversions of scientific minds, detailing the distinctive styles that define each individual researcher's approach to inquiry, as described by influential figures in the field such as Allen Newell, Werner Reichert, Herb Simon, and Raj Reddy.

  • Uploaded on | 0 Views
  • charlee charlee

About Scientific Lives: Desires, Diversions, and Styles

PowerPoint presentation about 'Scientific Lives: Desires, Diversions, and Styles'. This presentation describes the topic on This text explores the varied desires and diversions of scientific minds, detailing the distinctive styles that define each individual researcher's approach to inquiry, as described by influential figures in the field such as Allen Newell, Werner Reichert, Herb Simon, and Raj Reddy.. The key topics included in this slideshow are Scientific lives, Nomadic existence, Algorithmic complexity theorists, Architectures, Bounded rationality, Speech recognition,. Download this presentation absolutely free.

Presentation Transcript


1. Desires and Diversions Allen Newell Carnegie Mellon University Wed 4, Dec 91

5. Styles Of Scientic Lives Maxim : To each scientific life, its own style: And each style defines a life A nomadic existence Algorithmic complexity theorists A substantive theme Gordon Bell : Architectures, now multiprocessors A sequence of strategic objectives Werner Reichert of the Max Planck Institute, Tubingen A scientific idea takes five years to bring to fruition To work on interesting problems Their number is legion Whatever next seems publishable Publish or perish as the operative law of scientific life They shall be nameless A single ultimate scientific question Herb Simon : Bounded rationality Raj Reddy : Speech recognition Myself : The nature of the human mind

6. My Style : The Pursuit of a Single Desire The Nature of the Human Mind Maxim : Science is in the details What the question is How can mind occur in the physical universe? An answer must have the details : Generalities wont do What the question is not What is the nature of intelligence? (the AI question) How does bounded rationality explain? (the Simon question) The mind-body problem (the philosophy question) What is the nature of the brain? (the neuroscience question) An ultimate scientific question : A select company Why does the universe exist? When did the universe start? What is the nature of life? What is the origin of life? Why is evolution so stable? A grand unified theory of the universe To see an atom

7. When Does a Scientific Life Start? Maxim : The scientific problem chooses you, you dont choose it Not at age 17 : When I wanted to be a forest ranger Not at age 19 : When I wanted to be a scientist Not at age 22 : When I wanted to be an optical engineer Not at age 27 : When I wanted to be an organizational scientist But at age 27.7 : On a Fri afternoon in mid Nov 1954 A conversion experience : The Oliver Selfridge visit A pattern-recognition system on the Whirlwind MTC The insight : Computers can do any complex processing But understanding the mind was already determined This was just before the great burst of 1956-58 1956 : The Logic Theorist (LT) occurred 1956 : List processing occurred (IPL) 1957 : GPS occurred 1958 : The chess program occurred Nothing in 1954 presaged the imminent arrival of this burst

8. Diversions : And Their Uses #1 Maxim : Diversions occur, make them count Salvage what is possible for the main goal Gordon Bell and Computer Structures (1968-72) Diversion : To help a colleague write a book Uses : PMS and ISP were themselves a contribution Salvage : The concept of the architecture Raj Reddy and Speech Understanding Systems (1970-76) Diversion: Help the DARPA community and to help Raj Uses : Speech Understanding Program a major success Salvage: Only understanding of sensory processing

9. Diversions : And Their Uses #2 Stu Card, Tom Moran and The Psychology of HCI (1973-83) Diversion: Something on the West Coast Uses: Routine cognitive skill and GOMS, established HCI Salvage: Model Human processor as a unified theory Don McCracken, George Robertson and ZOG (1973-84) Diversion: Two great grad students and L* Uses: Helped start hypermedia, a major application Salvage: There are no short cuts to intelligence Points of note The diversions were all for social reasons They lasted a long time But always came back to the basic scientific goal

10. Failures : And Their Uses #1 Maxim : Embrace failure as part of success But use it for the main goal Merlin : With Jim Moore, Richard Young (1968-74) Idea : All Thought was seeing X as a Y, further specified Frames, attached procedures, general mapping, indefinite context dependence, automatic compilation Failure : We simply couldnt make it work Uses : It died and had no uses at all Did not learn that the idea was wrong Salvage : Need a global view, not a worms-eye view

11. Failures : And Their Uses #2 Instructable Production System Project (1976-1979) With Lanny Forgy, John McDermott, Mike Rychener, John Laird, Paul Rosenbloom Idea : Grow a production system by external means Obtain a large production system (1000 productions) Failure : Never got off the ground, never produced a thing Uses : Ops5 (Forgy), R1/XCON (McDermott), Universal Weak Method (Laird), Eng. AI (Rychener) Salvage : Launched Soar (Laird, Newell, Rosenbloom) Need default methods (universal weak method) Need organization of flat production systems Solution was problem spaces Points of note Failures can last a long time

12. Failures : And Their Uses #3 Chess language : With Bob Ramey (1962-63) L* : With Don McCracken, George Robertson (1969-74) The nature of design : With Peter Freeman (1969- 71) Automatic protocol analysis : With Don Waterman (1970-73) Problem generation : With Ramani (1973) Psychology of programming : With Reuven Brooks (1970-75)

13. Successes : And Abandoned Opportunity #1 Maxim : Solve whatever problems must be solved But do not be seduced by them List processing (1956-64) Ideas : Lists, dynamic resource allocation, date types, recursion, generators (streams) Abandon : Not participate in the list-processing literature Hashing : A dinner with Marvin Minsky (July 1961) Ideas : Associative access in random-access memories Abandon : Wrote one unpublished paper Protocol Analysis (1960-72) Ideas : Using sequential, NL content data for theory Re-establishment of introspection as legitimate Abandon : Not follow up with methodological studies But tried automatic protocol analysis (failed)

14. Successes : And Abandoned Opportunity #2 Efficiency of productions systems (1976-) Ideas : Rete algorithm, fine-grained parallelism, affect set, expensive chunks, unique attributes Not yet abandon : Because not yet usefully solved Points of note Success can last a long time too But never (ultimately) diverted A long life helps to copy with the diversions and seductions

15. Successes : And Their Cultivation Maxim : Preserve the insight and deepen it Deep ideas transform themselves beyond imagining Maxim : Deep scientific ideas are exceedingly simple Others usually see them as trivial List processing (1956) Search (1956) Symbols (1960) Problem spaces (1965) Weak methods (1969) Production systems (1972) Knowledge level (1980) Chunking (1981) Impasses (universal subgoaling) (1983)

16. A Deep Idea Transformed The Problem Space The idea : So trivial as to be uninteresting to most folk A problem space is a set of states and a set of operators from state to state. A problem is to go from the initial state to a desired state by applying a sequence of the operators. 1. The search space for heuristic combinatorial search (1957) 2. The space in which all problem solving occurs (1965) 3. The deliberately limited arena in which to solve a task (1957?) 4. The area in which all task accomplishment occurs (1979) 5. Multiple spaces formed by lack of knowledge (1983) 6. A least-commitment program control construct (1987) 7. A device for the formulation of tasks out of nothing (1990) 8. A device for general error recovery (199?) The problem space is all of these simultaneously

17. Deep Ideas Transformed Brief treatment of the others List processing (1956) Lists ->dynamic resource allocation ->attribute-values ->symbol systems ->production systems Search (1957) A method -> the method ->the weak methods ->UWM ->only one search method (MPD)(?) Symbols (1960) Problem space (1965) Weak methods (1969) weak methods ->search methods ->universal weak method -> problem spaces Production systems (1972) Operators ->recognition (association) memory Knowledge level (1980) A view ->operational concept Chunking (1981) Impasses (universal subgoaling) (1983)

18. Choosing the Final Project Maxim : Choose a final project to outlast you Maxim : Everything must wait until its time Science is the art of the possible Soar seems like a typical final project Look at what has to be done Have the right architecture Learn continuously from experience Communicate with the external world easily Learn continuously from its environment Live a long time Embody all its tasks simultaneously Become very large : 10^5 to10^6 associations Have a sense of history and place Have a system with a sense of self Learn from a social community It will surely outlast me! But Soar is not a final project : Simply a next project Chosen in 1983 (at 56) Choice was strategic : Aimed right at main goal Commitment comes from confluence of indications Problem spaces + productions, universal weak method, impasses, chunking, human cognitive architecture

19. Maxims for a Dedicated Scientific Life Maxim : To each scientific style, its own maxims Maxims occur throughout the talk: But not every one to work with results of field X, must be a professional in X Cognitive psychology, linguistics, neuropsychology We are all flawed : We all carry our burden of incapacities Know them like the back of your hand and dont fight them Mathematics as an example There is no substitute for working hard : very hard. The law of competition at the margin of excellence New things get started by evolution or chance, not design The next problem or what falls on the prepared mind A scientist is a transducer from nature to theory seek out nature and listen to her! My version of Herb Simons have a secret weapon The science is in the technique : All else is commentary What lives is what your scientific descendents must use to solve their problems Do your philosophy on the banquet circuit Symbol systems, Twenty questions, The knowledge level

20. Desires and Diversions Allen Newell 4 Dec 91 What happens with a whole scientific career? Mine isnt over, but its already forty years long. What shape and even purpose can be given to such a endeavor? Like all things human, the answers are diverse : there are many styles of scientific lives. The one I know best, of course, is my own. it hardly seems unique to me, but it does typify a definite style. I will tell some of the story of my own total scientific endeavor, and try to tell it to shed light on how to live a science.