3 Books By Andy Clark
Supersizing The Mind
PSU BD418.3.C532 2008 Oxford University Press
Too much "Scholar X says" for easy reading, but some good ideas from those scholars, which Clark amplifies. I skimmed, but would spend a week with this book if I had a week to spare.
Clark argues that if an external physical object performs a function that could conceivably be implemented as brain matter, it is effectively part of the EXTENDED mind. If Anne has a fully functioning memory, and Otto has a defective brain memory but a diary/book that he uses the same way, then his external and her internal memories are can be considered components of their respective minds.
Contrast to BRAINBOUND: human cognition depends directly on neural activity alone.
- p030 May 2004 Wired "Fear and Loathing on the Human-Machine Frontier" Bruce Sterling fears robot peripherals for the old. Clark claims this is a misconcieved vision of humans as "locked-in agents"
p033 Australian performer Stelarc with third robot hand, controlled by abdominal and leg muscle twitches. Has become "transparent equipment".
- Monkey with electrodes and robot arm, back and tongue imagers for the blind.
p071 epistemic actions: modifying the environment to extract information, i.e., opening the fridge to help plan dinner
p102 epistemic artifacts: diaries, compasses
p103 Sterelny perception operates in an environment of active sabotage by other agents
p103 change blindness and magic tricks
Brainfarts:
Politics and extended cognition. Individuals who control their immediate environment have better control of their cognition, minimizing sabotage by others. They control their own cogs, rather than be manipulated into cogs in a corporate or state machine. Downside: is littering a lazy way to mark the environment and leave a cognitive trail?
Penn Jilette is a magician, and exploits change blindness. He is also a libertarian, a political anti-magician.
Personal bicycles are tuned to the rider. Portland/Nike's orange "shared" bikes have some interesting technology, but are "one size fits all"; while that means that they all feel the same, they are not as optimal as a personal bike. So there a political statement here: "you are an interchangable cog", and those who don't fit the bikes are disadvantaged.
Music - Does the beat of tapping toes and other rhythmic body movements exploit mechanical resonances? Are musician preferences for particular piano actions because their skills require them? Can the piano-actions be precisely electromechanically mimicked? Can the adaptive piano-actions be used to train musicians, or make a great musician out of a merely adequate one?
Natural Born Cyborgs
Microcognition
PSU BF311.C54 1989
Clark questions the applicability of propositional, symbol-processing, rules-based AI
p5 The syntactic challenge demands that if beliefs are real and cause behavior, there must be neat, in-the-head syntactic analogues to the semantic expressions in sentences describing them. This is classical cognitivism. Clark denies this, calling belief and desire talk a net over the behavior of an embodied being acting in the physical world.
The system uses slow analog physical processes in parallel to quickly discover good-enough solutions to problems.
p9 Classical cognitivism and later conventional AI relates mind with computational modelling
p11 SPSS Strong Physical Symbol System hypothesis: A virtual machine engaging in the von Neumann-style manipulation of standard symbolic atoms has the direct and necessary and sufficient means for general intelligent action.
p14 BACON Newell and Simon 1979, serial heuristic search (out of a small subset) for "scientific laws" that match data
- p15 Contrary to Simon, Hofstadter asserts that everything of interest in cognition occurs below the 100 msec level
- p16 flash of insight, not sequential-methodological
p20 STS Syntactically Transparent System, structured descriptions representable by symbols, processable by other symbols
KHL are quantized systems, representable by finite fields of finite numbers and computable with discrete operations, considered "symbolic" by STS advocates, even though the symbolic representation is fantastically larger and slower than the object modelled?
p21 Functionalism symbol manipulation matters, not the material
- Hilary Putnam (1960, 1967), Jery Fodor (1968) , David Armstrong (1970), Daniel Dennett (1981) and William Lycan (1981).
p22 psychological / mental state p23 mediates i/o and affects other internal states .p24 no qualia raw feels, real subjectivity
p66 Ask not what's inside your head but rather what your head is inside of Mace, 1977
- p66 In 1962, Herbert Simon wrote about complex systems evolving from simpler useful components.
p73 functional philology of mind list of things evolved brains do
p84 Hebbs Rule if two units are simultaneously excited, increase the connection strength between them
p85 Rosenblatt's perceptron overhyped
PDP Parallel Distributed Processing, Connectionist AI
- no explicit schemata, they emerge from from parallel network activity of simple elements
McClelland and Rumelhart
- interference between similar patterns creates PDP mistakes, similar to human mistakes
NO TIME TO FINISH THIS
SHRDLU
SPSS Strong Physical Symbol System: a symbol-manipulation machine has the necessary and sufficient means for intelligent action.e
p20 STS Syntactically Transparent System
https://www.newyorker.com/magazine/2018/04/02/the-mind-expanding-ideas-of-andy-clark