1. Sprevak and Colombo. INTRODUCTION. A bird’s eye view of the handbook. Charting the history of computational mind, from Turing through Fodor, Pylyshyn, PDP and dynamical systems. Now machine learning and deep learning. What are the commonalities, differences and challenges?
There will be four sections: History and Future, Types of Computing, Foundations and Challenges, and Applications.
Already intrigued by the tension between Fodorian linguistic approaches to CTM and those illuminated by RL, ML, and Deep Learning. Cued to challenges of implementation and scalability - cf the Frame Problem - how do we know/compute when an event has occurred?
Beyond Simon’s (1957) Bounded Rationality, how with limited time, memory and attention do we/computers solve complex, pressing, and ambiguous problems in the real world?
They hint that minds, brains, brain regions, circuits needn’t have a single calculus...computational scientists can be pluralists, picking methods and approaches for specific problems. I like the idea that minds and brains/regions might do so too.
And I’ve already failed at one tweet per chapter. This is going to be hard. But it’s testimony to my interest in the topic and the quality of the writing
1 (proper). Computational Thought. Alistair Isaac. On the metaphors for mind being influenced by contemporary technology. For Descartes, rational soul observed the flow of fluid through pipes, as a king views his garden. Without soul no novel behavior possible.
Vs. Turing (and Lovelace) who worried that mechanistic computations May so outstrip human prediction to produce actions that appeared novel (but weren’t) - thus the test
La Mettrie in ‘Man, a Machine’ - muses that language in humans could be built from simpler mechanisms in animals. And that vibrations may be key. As neuroanatomy stuttered, such thinking did too. Until Ramon Y Cajal and the neuron doctrine
Leibniz and Boole - on computations the mind might make, for L - the representations are propositions: subject and predicate, and they are manipulated with a logic of discovery (so novel propositions can be surmised). B - emphasizes true and false, 1 and 0, preempting Shannon
More general purpose machines - which worked differently depending on a program - were demanded by the textile industry in c18th. The Jacquard Loom led to Babbage’s analytical engine
Lovelace cautioned against computationalism - we should neither under rate nor overvalue the capacities of the analytical engine (this seems even more apt today)
Turing called this Lady Lovelace’s Objection (which would be a great band name).
In sum, Isaac’s chapter emphasizes mechanism and thought as computation - I had heard the fountain analogy. I loved learning about the loom as inspiration for the analytical engine
Next, 2. Proudfoot and Copeland - Turing and the first electronic brains. How news of the first computers broke and the debate sparked: where they simply adding machines or capable of intelligence and relevant to mind?
(I need to stop for tonight - but I love this book so far. The chapters are pitched just right and are a perfect length - finding it hard to put down)
Mountbatten announcing Turing’s war time achievements: ‘possible to evolve an electronic brain...a machine that can link information it receives about machinery under it’s control and action needed to keep it working as desired...judgment...choice...a mediocre game of chess!’
Hartree detracted - they are only capable of doing what they are instructed, triggering claims that computing machines could at best imitate the processes of the unconscious brain...the new Frankenstein...the big brother of the sliderule
Turing countered that the machines would learn from experience and human contact, with enough memory
In Intelligent Machines he describes the first neural networks, trained to perform specific tasks (this was not implemented until 1954 at MIT)
1952, Guardian: a machine was being developed that could examine itself for defects when it made errors and thus develop a better electronic brain
The debate focused on what it means to think and whether the phenomenology or thinking distracts us from allowing that machines can think. Turing shoes away from defining thinking, arguing instead that by studying thinking machines we’d learn how we ourselves think
Chapter 3. British Cybernetics. Joe Dewhurst. TIL of the Ratio club (incl Walter, Ashby and Turing) comprised of those who had Weiner’s idea (cybernetics) before Weiner. They were interested in embodiment (w actual bodies, not just physical implementation)
Walter built a robo-tortoise with simple sensors. It showed apparently intentional behavior - like seeking light - much like one of Braitenburg’s Vehicles
Ashby built a homeostat to demonstrate the impact of linked sensors arriving at a common set-point 
They influenced enactivism and autopoesis. Whilst some enactivists are anti-computational, the chapter concludes that this is due to a narrow (semantic) view of computation. As we know, @micahgallen, embodiment, computation and cybernetics are entirely compatible
Other cool tidbits - Ashby had an interest in psychiatry and R.D. Laing apparently a cyberneticist - early British Computational Psychiatry FTW
Walter wanted to break his robo-tortoise to gain insights into neurological and psychiatric illness
(Oddly watching robocop while reading about cybernetics)
(Off topic, though also very on topic, this is a great movie)
4. Cybernetics by Tara Abraham. McCullough, Pitts, Von Neumann - artificial neurons in networks. On how well they model the brain, esp when they are digital - which for computers was a huge advance but for neurophys...it’s more complicated - particularly wrt neuromodulators
(Excited Von Neumann is in the mix - I love ‘The Computer and the Brain’ - it was to be his Silliman Lecture at Yale had he not passed away. In it he apologizes for not being a neuro/psych yet speculating on the brain. He really wrestles with the fit btwn computers and brains)
McCullough was also a neuropsychiatrist and Pitts a math prodigy, inspired by Leibniz. Read their sad tale here: m.nautil.us/issue/21/infor…
Oppenheim and Putnam believed McCullough, Pitts, Von Neumann provided the sort of reductionism necessary for unity of knowledge (consilience of brain and mind). For McCullough in particular, empirical data challenged his models (incl Karl Lashley on cortex).
McCullough it seems wanted his models to tested with data. To inspire neuro and psychiatric studies. Letvinn said “he would rather have been clearly wrong than maunderingky vague” - something for us all to aspire to
5. Aizawa- the ‘conception’ of cognitive science: revisit 3 foundational papers from CTM and discern what they really said abt Turing equivalence. Miller’s 7+/-2 paper, Chomsky on grammars and the structure of linguistic theory and Newell & Simon’s Logic Theory Machine...
In each case, Aizawa finds the allusions to Turing and completeness were mostly in passing - unlike how those papers are referred to today
For each, computers were a tool or a loose metaphor. Turing-equivalent computation played a secondary role at most
(One theme I hope develops, the recent shift toward reinforcement learning and deep-learning and away from semantic representations: What does it mean for anti-computation and the (imho erroneous) claim that Chomsky killed behaviorism chomsky.info/1967____/)
