structure for 'Thought'    |     alphabetical list of themes    |     unexpand these ideas

---

18. Thought / B. Mechanics of Thought / 6. Artificial Thought / a. Artificial Intelligence

[possibility of building thinking/conscious machine]

10 ideas

7651	With wonderful new machines being made, a speaking machine no longer seems impossible [La Mettrie]
	Full Idea: If wonderful machines like Huygens's planetary clock can be made, it would take even more cogs and springs to make a speaking machine, which can no longer be considered impossible, particularly at the hands of a new Prometheus.
	From: Julien Offray de La Mettrie (Machine Man [1747], p.34)
	A reaction: Compare Descartes in Idea 3614. The idea of artificial intelligence does not arise with the advent of computers; it follows naturally from the materialist view of the mind, along with a bit of ambition to build complex machines.

8056	AI can't predict innovation, or consequences, or external relations, or external events [MacIntyre]
	Full Idea: AI machines have four types of unpredictability: they can't predict radical innovation or future maths proofs; they couldn't predict the outcome of their own decisions; their relations with other computers would be a game-theory tangle; and power failure.
	From: Alasdair MacIntyre (After Virtue: a Study in Moral Theory [1981], Ch. 8)
	A reaction: This isn't an assertion that they lack 'free will', just a very accurate observation of how the super new machines would face exactly the same problems that we ourselves face.

7654	What matters about neuro-science is the discovery of the functional role of the chemistry [Dennett]
	Full Idea: Neuro-science matters because - and only because - we have discovered that the many different neuromodulators and other chemical messengers that diffuse throughout the brain have functional roles that make important differences.
	From: Daniel C. Dennett (Sweet Dreams [2005], Ch.1)
	A reaction: I agree with Dennett that this is the true ground for pessimism about spectacular breakthroughs in artificial intelligence, rather than abstract concerns about irreducible features of the mind like 'qualia' and 'rationality'.

12655	Frame Problem: how to eliminate most beliefs as irrelevant, without searching them? [Fodor]
	Full Idea: The frame problem is, precisely: How does one know that none of one's beliefs about Jupiter are germane to the current question, without having to recall and search one's beliefs about Jupiter?
	From: Jerry A. Fodor (LOT 2 [2008], Ch.4.4)
	A reaction: Presumably good chess-playing computers have made some progress with this problem. The only answer, as far as I can see, is that brains have a lot in common with relational databases. The mind is structured around a relevance-pattern.

3135	Is thought a syntactic computation using representations? [Fodor, by Rey]
	Full Idea: The modest mentalism of the Computational/Representational Theory of Thought (CRTT), associated with Fodor, says mental processes are computational, defined over syntactically specified entities, and these entities represent the world (are also semantic).
	From: report of Jerry A. Fodor (works [1986]) by Georges Rey - Contemporary Philosophy of Mind Int.3
	A reaction: This seems to imply that if you built a machine that did all these things, it would become conscious, which sounds unlikely. Do footprints 'represent' feet, or does representation need prior consciousness?

3215	Images can't replace computation, as they need it [Rey]
	Full Idea: Processing of images and mental models seems to require, and therefore is unlikely to replace, computation and representation.
	From: Georges Rey (Contemporary Philosophy of Mind [1997], 10.1.2)
	A reaction: A good point. If you are a fan of mental imagery, you still have to explain how we can hold an image, or recall it, or manipulate it. I always, I don't know why, wince at the thought of 'computations' among neurons.

3194	CRTT is good on deduction, but not so hot on induction, abduction and practical reason [Rey]
	Full Idea: The computational/representational theory of thought has given a good account of deduction, but mechanical theories of induction, abduction and practical reason are needed in order to make a machine which could reason.
	From: Georges Rey (Contemporary Philosophy of Mind [1997], 8.5)
	A reaction: This is the best analysis of rationality that I have found (four components: deduction, induction, abduction, practical reason). I can think of nothing to add, and certainly none of these should be omitted.

6655	The 'Frame Problem' is how to program the appropriate application of general knowledge [Lowe]
	Full Idea: The 'Frame Problem' in artificial intelligence is how to write a program which not only embodies people's general knowledge, but specifies how that knowledge is to be applied appropriately, when circumstances can't be specified in advance.
	From: E.J. Lowe (Introduction to the Philosophy of Mind [2000], Ch. 8)
	A reaction: As Lowe observes, this is a problem, but not necessarily an impossibility. There should be a way to symbolically map the concepts of knowledge onto the concepts of perception, just as we must do.

6657	Computers can't be rational, because they lack motivation and curiosity [Lowe]
	Full Idea: Lack of motivation and curiosity are perhaps the most fundamental reason for denying that computers could be, in any literal sense, rational beings.
	From: E.J. Lowe (Introduction to the Philosophy of Mind [2000], Ch. 9)
	A reaction: I don't see why programmers couldn't move those two priorities to the top of the list in the program. When you switch on a robot, its first words could be 'Teach me something!', or 'Let's do something interesting!' Every piece of software has priorities.

23518	Modern AI is mostly machine-based pattern recognition [Seth]
	Full Idea: Much of today's AI is best described as sophisticated machine-based pattern recognition.
	From: Anil Seth (Being You [2021], IV.13)
	A reaction: Personally I wouldn't want to underestimate the extent to which human intelligence is also pattern recognition (across time as well as in space).