Views on Turing's paper on Computing Machinery and Intelligence

The paper on Computing Machinery and Intelligence by Alan Turing is probably what ignited study and research in the fields today allied to Artificial Intelligence. In section 7 of this seminal paper, he talks about a variety of aspects which he considers most useful when creating a machine which should be able to learn. 

Starting from the analysis and interpretation of ideas, the mechanics involved in information storage and retrieval, rules for creating such a machine, and on to finer aspects of how it can be more human-like in its processes, Turing takes us through a series of arguments and statements which he defends and elaborates on, with respect to other such ideas given in previous sections. 
In my opinion, there are certain fallacies in the reasoning and places where the offered evidence is not concrete enough to support the overlying ideas.

In the beginning, Lady Ada Lovelace's objection on a machine not being able to "originate" any material is cited as a starting point on a discussion on the extent of what a machine can do. Turing speaks about injecting an idea into the machine, where it could hang in quiescence after an initial response takes place, and goes on to talk about supercritical and subcritical minds. He however fails to mention the exact process here which could be equated to learning. The aforementioned response may well be an automated one, due to the application of certain stimuli to the machine. A line of reasoning worth following is isolating the point at which the machine is able to gauge (if at all) whether or not the external stimulus is worth responding to.

Moving on to the topic of information storage in the brain, Turing assumes that the problem of developing a machine which can learn is that of writing the perfect program for it. While that may be so, his further assumptions about the extent of storage in the human brain (which are figures far far below what the actual ones must be) are totally fallacious! Given that the paper was written much before present technology to get an accurate idea of the above was developed, his belief that the higher types of learning occupied only a small fraction of the brain is totally unsupported by evidence, and is unconfirmed even today.

As he proceeds further, Turing mentions that an important feature of a learning machine is the teacher being in ignorance of the actual working of the machine's brain (or whatever acts closest to it). There is a missing piece of information here - are all the teachers people not involved in the actual development of the machine itself? Doing so would be a disadvantage, since the possibility of seeing a brain (no matter how basic) actually undergo the process of learning would give us insights never before imagined, into the human thinking process. A better way to approach this would be to give some teachers full knowledge of the brain's workings, and the option of tailoring their inputs to suit the machine itself. Comparing the two forms of education and fine tuning our methods would enable us to glean a lot more information about the entire process.

Before concluding, Turing introduces the concept of a random element in the learning process, by stating an example in which the solution to a problem is being searched. Instead of it being done sequentially, he proposes numbers be chosen randomly till the required solutions are found. I think that this idea would have been better utilized if the random element is introduced with respect to the kind of response which the machine could have to certain types of input. Human reactions tend to be spontaneous, and such randomness is inherent to us - certain people react differently to different sets of stimuli at different points, and this is what we would want the machine to emulate, so as to be able to get as close to simulating the human brain as possible.

The computability of all brain processes (being possible or not) is a topic still under fierce debate today, and we see that this was a fundamental assumption when this paper was written. It would be instructive to see whether or not our lines of thought today are different from those under consideration then, and either way, see if theories being put forth today may someday appear incredulous 20 years down the line.


Resources:
Abelard's notes on Turing's paper