PHL309 Logic, Language, and Thought Professor: Craig DeLancey
Office: CC212C
Email: delancey@oswego.edu



Past Assignments
30 January
Read the selection from Aristotle. Come to class prepared to explain what Aristotle's view on the infinite.

1 February
Get a head start and begin reading the selection from Galileo's Two New Sciences: First Day.

I emailed you the prefered translation. If you didn't get that email let me know.

But, if that isn't up yet, an fair translation is freely available here.

Our goal is to ultimately read up to but not past the passage by Simplicio that in this translation begins "I cannot help admiring your discussion; but I fear that this parallelism...." In the version I'm trying to get back onto e-reserves it is page 44 (or, original edition page numbers are on the side margins, and by those we would like to get to 82.) You're welcome to read farther, but we won't need to discuss past this point and I won't require that anyone read past this point.

4 February
Read from the Galileo selection. You are welcome to start at the beginning if you can, but I only need you to read from page 35 to 42. I think there are at least five arguments that we cannot reason about infinity. Try to identify them, and see if you can understand them. We'll go over the first few. I will ask you to explain the last several.

6 February
Homework. On page 40 (editor's number 78), Salviato explains a paradox having to do with squares and roots of numbers. Write out this argument as a reductio argument. This means: What is the conclusion? What is the assumption for reductio? What are the necessary steps of the argument? What is the contradiction that Galileo finds? (After Monday, we'll have several examples of doing this, so you'll have a model to copy.) You should be able to do this on a page, or at most two. Hand it in at the beginning of class.

Several people asked me for a model. Here is how we reconstructed the soupdish argument in class.

Galileo wants to show we cannot reason about actual infinities, so he sets out to show if we could then we would contradict ourselves.
  1. Assumption for reductio ad absurdum: We can reason about actual infinities.
  2. If we can reason about actual infinities, then reason tells us that the following are true:
  3. There is a geometric figure, called the soupdish, such that a planar cut perpendicular to the axis yields two figures of the same volume (a cone and a ring).
  4. Volume is composed of quantity of points, so if two figures have the same volume then they have the same number of points.
  5. From steps 3 and 4, we know that the two figures from some cut across the soupdish yield a ring and cone that have the same number of points.
  6. The last cut across the soupdish yields a circle and a point.
  7. Steps 3 and 6 tells us that the circle and point have the same volume.
  8. A circle has infinitely many points. A point has one point.
  9. Step 8 tells us that the two figures have different number of points.
  10. Step 7 and 4 tell us that the circle and point have the same number of points.
  11. Steps 9 and 10 are a contradiction. [This means we put something false into our argument. Galileo concludes that the false thing we put in was:]
  12. Step 1, the assumption for reductio, is the false claim that caused the contradiction. Therefore: we cannot reason about actual infinities.
Your argument will likely start in a similar way (steps 1 and 2 can be the same), but you must identify the explicit contradiction (some statement and also the denial of that statement).

Here's how I graded this: 1 point for identifying a plausible conclusion; 1 point for identifying the assumption for reductio (this is just the denial of the conclusion, so it's a freebie); 2 points to identifying the essential steps of the argument; 2 points for explicitly identifying the contradiction of the argument.

February 8
We'll look at Galileo's version of how we can reason about something like potential infinities; this is at the end of our selection.

Remember that this day I cannot have office hours. Sorry.

February 11, 13
A basic introduction to set theory! And cool facts about Cantor.

February 15
Howework. Answer the following questions. Work on your own, since the point here is to determine where you still need work on set theory. Write up your answers and bring them to class.
  1. Which of these are true?
    a. {1} ∈ {1, 2, 3}
    b. 2 ∈ {1, 2, 3}
    c. {1} ∈ {{1}, {2}, {3}}
    d. 1 ∈ {{1}, {2}, {3}}
  2. Which of these are identical?
    a. {1} and {{1}}
    b. {2, 3, 1} and {1, 2, 3}
    c. {} and {{}}
    d. {7, 7} and {7}
  3. Give an example of a proper subset of each of the following sets.
    a. {1, 2, 3}
    b. {a, b}
    c. {{1}, {2}, {3}}
  4. For each of the following sets, list all of its subsets. (Remember: if we do not require that the subsets be proper, you can include as a subset the set itself.)
    a. {1}
    b. {1, 2, 3}
    c. {}
    c. {{}}
  5. Assume these sets continue in the simplest way as listed. For problems a and b, can you give an example of a function that relates the first to the second in a 1-to-1 correspondence (an on and onto function that would also be a function if it went backwards -- what mathematicians call an inverse)? For problem c, can you find a function that is on and onto, even if it were not 1-to-1? (You can answer these using some basic arithmetic -- no need to get fancy; for example, the function relating the naturals to the even numbers would look like: f(x) = 2x.)
    a. {2, 4, 6, 8, 10....} and {20, 40, 60, 80, 100....}
    b. {1, 2, 3, 4, 5....} and {1, 3, 5, 7, 9, ....}
    c. {1, 2, 3, 4, 5....} and {0}
    (Note that for c, the second set contains only one element, namely the number zero.)
  6. Opinion question: if Cantor is right that sets of infinite cardinality can have infinite proper subsets, then does it follow that all infinite sets are the same size?

February 20
We will review Cantor's claim; review in a more formal way Cantor's diagonal argument; and introduce Cantor's Theorem.

February 22
A review of functions; a review of powersets; and then introduction of Cantor's theorem. We can do a fun proof of this, before we state the proof more formally.

Also: There were typos and a mistake in my handout on set theory, so I rewrote it. And there's a bonus! I added a little set of drawings that I think make 1-to-1 more clear than I was making it before (quite a few times I mistated it). I'll bring this to class, but I'll also post it here.

February 25
Homework due at the beginning of class.

These are some more problems meant to help us think more about sets and powersets. Write up your answers to these and bring them to class to hand in.

  1. What are the powersets of the following sets?
    • {1, 2}
    • {}
    • {a, b, c, d}
    • { {a, b}, {{7}} }
  2. What are the following sets powersets of? That is, the powerset of what set would be the following?
    • {{}, {a}, {b}, {a, b}}
    • {} [<-- Typo! I meant to write: {{}} ]
    • {{}, {1}}
    • {{2}, {4}, {6}, {4, 6}, {2, 4}, {6, 2}, {}, {2, 4, 6}}
  3. Which of these sets has the largest cardinality? Explain what this means and how you know your answer is correct. You'll probably need to define cardinality, and describe a one-to-one function, and then apply these notions.
    • The natural numbers, N
    • The even numbers (you can call this E)
  4. Simplicio argued that a line segment AB that is twice as long as a line segment CD has twice as many points; but also AB and CD have infinitely many points, and infinitely many points is as many as infinitely many points, so AB and CD have the same number of points. Thus, the segments have differenly quantities of points (AB has twice as many points as CD) and the same quantity of points (AB and CD have the same number of points -- namely infinitely many). If we agree with Cantor, how many points are on the line segment CD? How many are on the line segment AB if it is twice as long as CD? What does this tell us about Simplicio's argument? Does Cantor's Claim matter here? (Hint: in modern geometry, every line segment has the same cardinality of points -- namely, aleph1.)
Meanwhile, in class, we will review Cantor's theorem, and then discuss what this means for the infinities.

27 February
In class:Review of the homework. Review of all of our set theory. We will finally begin to answer directly each of Galileo's objections to reasoning about infinity!

Regarding last class: we reviewed in the last class a reductio proof of Cantor's Theorem. I described a hopefully-easier-to-understand case for N and R, but promised to write out a more general proof. A more general proof would look like this.
Cantor's Theorem says that, for any set S, |P(S)| > |S|.

We can easily see that |S| ≤ |P(S)|, because there is a function from P(S) and onto S. This function has as its domain the singleton members of P(S), and maps each of these onto (or, if you prefer, points each of these at) its member in S. That is, {x} in P(S) will be mapped onto x member of S. (Look back at your handout: you'll see that if we have a function from a set A and onto a set B, then we know at least that |A| > |B|; that's the principle I used here.)

So, since |S| ≤ |P(S)|, is |S| < |P(S)|, or is |S| = |P(S)|? We rule out the latter case.

Suppose |S| = |P(S)|. Then, there is a function F on S and onto P(S), and this function is 1-to-1. Let F be such a function. We claim that if F existed we would get a contradiction.

Let NPS be the set of all members of S such that x is not the output of F(x). This is the set of all those elements of S that when put into the function F, put out some member of P(S) that the member does not belong to (I chose the long name NPS to helps us think, Not-Pointing-at-a-Subset-we-belong-to). Our fancy way of writing this with set theory notation is:

NPS = {x ∈ S | x ∉ F(x) }

Note that NPS must be in P(S), since P(S) contains every possible subset of S.

But now we have a problem. What member of S is mapped to NPS? Call this m. Is m ∈ NPS? If m is, then m is in the set of elements of S that do not point at a set that they are in; so that contradicts that F(m) = NPS (to say that F(m) = NPS is just to say F maps m to NPS, or that F makes m point at NPS). So consider the only other alternative: m ∉ NPS. But since F(m) = NPS, this means m points at a set that m does not belong to; and so by definition m should be in NPS!

Either option yields a contradiction. The source of our contradiction, we conclude, is the assumption that |S| = |P(S)|. So, since we already knew that |S| ≤ |P(S)|, we conclude |P(S)| > |S|.

This proof is general. We said nothing about the particular contents of S.

1 March
Continuing to review Galileo's objections to reasoning about infinity, and Cantorian replies.

I promised to put my reconstruction of Galileo's arguments here.

Galileo's Reductio Arguments (DeLancey's Interpretation)

I. The soupdish.

1. We cannot reason about actual infinities.
2. For (assumption for reductio ad absurdum): suppose we could reason about actual infinities.
3. If we could reason about actual infinities, then reason would tell us:
4. There is a geometric figure which we have called "the soupdish": every cut with a plane above the intersection of the cone and dish perpendicular to the axis yield a ring and cone of the same volume.
5. The last cut of the soupdish yields a point and a circle.
6. A point has one point, a circle has infinitely many points, so the circle has greater volume than the point (and so the circle is not the same volume as the point).
7. But 4 and 5 tell us that the point and circle have the same volume.


II. Composition of indivisibles.

1. We cannot reason about actual infinitesimals.
2. For (assumption for reductio ad absurdum): suppose we could reason about actual infinitesimals.
3. If we could reason about actual infinitesimals, then reason would tell us:
4. Every line can be cut in two equal halves.
5. Three points (or any odd number of points) together would form a line.
6. By 4 and 5, we could divide a line composed of an odd number of points in half.
7. But that means we must cut a point in half, since an odd number of points will have a point at its center.
8. But by definition we cannot cut a point in half.

III. Comparing sizes of infinities on lines.

1. We cannot reason about actual infinities.
2. For (assumption for reductio ad absurdum): suppose we could reason about actual infinities.
3. If we could reason about actual infinities, then reason would tell us:
4. There are lines which are longer than other lines (for example, suppose AB is half as long as a line ABC that contains AB).
5. Each line has infinitely many points (for example, AB and ABC both have infinitely many points).
6. A line twice as long as another line has twice as many points (for example, AB has half as many points as ABC); so the two lines have a different number of points.
7. But by 5, any two lines have the same number of points: namely, infinitely many (so AB has infinitely many points, and ABC has infinitely many points).

IV. Comparing naturals and squares: correspondence argument.

1. We cannot reason about actual infinities.
2. For (assumption for reductio ad absurdum): suppose we could reason about actual infinities.
3. If we could reason about actual infinities, then reason would tell us:
4. We can identify a 1-to-1 function on the natural numbers and onto the squares of the natural numbers.
5. From 4, we conclude there are exactly as many natural numbers as squares of natural numbers.
6. The squares of the natural numbers are a proper subset of the natural numbers.
7. From 6 we conclude that the squares of the natural numbers is smaller than the natural numbers (and so there are not exactly as many natural numbers as squares of the natural numbers).

V. Comparing naturals and squares: growth argument.

1. We cannot reason about actual infinities.
2. For (assumption for reductio ad absurdum): suppose we could reason about actual infinities.
3. If we could reason about actual infinities, then reason would tell us:
4. We can identify a 1-to-1 function on the natural numbers and onto the squares of the natural numbers.
5. From 4, we conclude there are exactly as many natural numbers as squares of natural numbers.
6. From 5, we conclude that the quantity of naturals and squares grow towards infinite quantities at the same rate as we count through the numbers.
7. But for each square matched to a natural number, as we count through the numbers, more and more natural numbers are left out of the set of squares.
8. From 7, we conclude that the quantities of naturals and squares do not grow towards infinite

Our task was to see which of the premises Cantor and those who agree with Cantor would reject, if we do not agree that the problem is the assumption for reductio. Keep in mind two principles of modern geometry: that all geometric figures (other than a point) have aleph-1 many points; and that measures like length, area, and volume are not determined by the quantities of points (all lengths, areas, and volumes have aleph-1 points). Combine those observations with Cantor's Claim, and it becomes clear what Cantor would reject in each argument.

4 March
In class: Some review. Then, a question: when do we know if a set exists? Consider the special case of the set of everything. Does that exist?

Optional homework: The last two problems on the last homework were harder than I expected, and I did not phrase them as easily as I could have. So here's a chance to do them over. If you do this, just staple this to your old homework. I'll just replace those points on your last homework. (E.g., these 7 points will replace those 7 points on the last homework.)
  1. What are the comparative cardinalities of these two sets (is one larger than the other, or are they the same size)?
    • The natural numbers, N
    • The odd numbers (you can call this O)
    If you say they are the same size, identify a 1-to-1 function on N and onto O. If you say they are different sizes, explain why you think they are (for example, is one the power set of the other?).
  2. Simplicio (one of the character's in Galileo's dialogue) argued that a line segment AB that is twice as long as a line segment CD has twice as many points; but also AB and CD have infinitely many points, and infinitely many points is as many as infinitely many points, so AB and CD have the same number of points. Thus, the segments have differenly quantities of points (AB has twice as many points as CD) and the same quantity of points (AB and CD have the same number of points -- namely infinitely many). This contradiction led him to conclude we cannot reason about actual infinities.

    However, if we agree with Cantor, there is no contradiction here (and Simplicio/Galileo made a mistake in that argument above). For Cantor: How many points are on the line segment CD? How many are on the line segment AB if it is twice as long as CD? What does this tell us about Simplicio's argument?

6 March
Taking stock of where we are at. Maybe some set theory review and practice, including the set-theory antinomy. Review of where we've come to -- and how should we respond to the set-theory antinomy?

I will answer any questions you might have about the test.

Frege! Logicism!


11 March
We can review anything you would like to review in preparation for the test, but you have to bring questions to class in order for us to be able to do that. If that doesn't take all our time, we'll return to Frege.

13 March
Quiz on: basic concepts of set theory; reconstructing Cantor's diagonal argument; reconstructing Galileo's arguments that we cannot reason about infinity; Cantorian replies to Galileo's arguments; Cantor's Claim, and how it answers some of Galileo's challenges; Cantor's Theorem, and how it led to the set-theory antinomy.

I'll include with your test the handout on sets, so that you have with you the definitions we've been given.

15 March
We will review the test. The average (mean) was 28; but the average (mean) for those who skipped one or more homeworks was 19. I've given you all midterm grades, which I assume are available online. Those are distorted because they count a few homeworks as half your grade. But we'll strive to have at least five more homeworks, in order to bring more balance to that. The midterm grades are approximate.

I can share later my grading key.

After our review, time allowing, we'll talk about Russell's postcard to Frege.

25 March
Hilbert's problems!

27 March
We will review the concepts of: consistency, decidability, completeness. We will also discuss a brief detour into intuitionism. And then: more paradoxes!

If you purchased Logicomix, now is a good time to read the first 3 sections. This would take you to page 154. There are many historical inaccuracies in here, which the authors took in order to simplify and dramatize the story. The basic thrust is not inaccurate, however; so don't treat it as a primary source but rather as a dramatic introduction to the portion of our potted history that we are discussing now. It's a fun book.

Some of you have asked me if there are other classes on this kind of stuff. If you like logic or these kinds of weird discoveries (such as those about infinity), you might want to consider a minor in logic. The requirements for the minor are available here.

1 April
Principia Mathematica!

If you purchased Logicomix, finish it.

3 April
Introduction to Godel.

If you got the book by Casti, Godel, read chapters 1-4. Additional resources include the Stanford Encyclopedia entry on Godel; this is often technical, but the first few sections provide some accessible background.

5 April
If you chose to buy Godel, now is a good time to read (up to and including) chapters 6 and 7.

8 April
Homework due. I'll bring the homework to class on 3 April, but also you can download it here.

10 April
We'll review again some questions that have come up about Godel; then return to Turing's brilliant idea and get ourselves comfortable with it.

It can be helpful to use a Turing machine simulation, to play with a machin and get a feel for it. Here's one that may work for you: http://math.hws.edu/TMCM/java/labs/xTuringMachineLab.html. You could test your machine on it. You'll have to turn on Java in your browser for it to work.

The Wikipedia entry on Turing machines is actually quite thorough, if you want another resource. The Stanford Philosophy Encyclopedia is more technical, and specifies the machine slightly differently.

12 April
We'll review some more machines, and introduce the idea of a universal Turing machine.

15 April
Homework due. Working alone, you will write descriptions for 2 turing machines. You will need to provide for each: the rule table, a description of how the tape will be interpreted and configured (including how to interpret the output), and where the head starts on the tape.

The machines should not be universal!

It can be helpful to use a Turing machine simulation. Here's one that may work for you: http://math.hws.edu/TMCM/java/labs/xTuringMachineLab.html. You could test your machine on it.

Try to do the machines with only 1s, 0s (or 1s and blanks), and as few other characters as you need as your alphabet (sometimes it makes things a lot easier to add a character, for example to mark the end of a number, so do that if it helps you!).

The machines should be:
  1. A machine that tells us whether a number is odd or even. (You can write a number n as n "1"s on the tape; that's the easiest way to do it; then your machine would tell is if the number is odd or even.)
  2. An addition machine for numbers n, m where always m > 0 but n could be 0. (You can write a number n as n "1"s on the tape and m as m "1"s on the tape; that's the easiest way to do it. Consider always starting with n.)
For each problem, clearly describe the proper input and its interpretation, and the expected output and its interpretation. That is, where does the tape head start and finish in relation to the data, the answer, etc.

In class this day, we'll review the idea of a UTM, and then return to Hilbert's Question about effective procedures. Have we now answered that question?

17 April
I'll have office hours 1-2, trying to fit to Quest's schedule. I hope that works for you.

19 April
We will review our homework; review Turing's diagonal argument that we saw on Monday; then we'll look at the crazy machine version of the Halting Problem.

22 April
We will review the Halting Problem. Then, our topic will be: a surprising observation: what the Halting Problem Result tells us about prediction.

Before class, play with a version of the LifeWorld; a good one can be found here. Can you identify a structure that stays the same over time? Can you identify a structure that "moves" across the world?

24 April
Homework due. You may work in teams; write the name of every team member on the homework. Due at the beginning of class. You are going to make two Turing machines (none universal!) and write up their specification. You'll hand in that specification.

It will be very helpful to use a Turing machine simulation. Here's one that may work for you: http://math.hws.edu/TMCM/java/labs/xTuringMachineLab.html.

Try to do the machines with only 1s and 0s (or 1s and blanks), and as few other characters as you need as your alphabet (sometimes it makes things a lot easier to add a character, for example to mark the end of a number, so do that if it helps you!).

The machines should be:
  1. an addition machine; it should be able to handle 0+0 or 0+ any number or any number + 0. (If you are using 1s and blanks, then a zero may be a blank; and 0+5 might be written on the tape as (starting leftmost under the tape head) "blank blank 11111". That is, the first blank is zero, the second is interpreted as separating your two numbers. 0+0 would be "blank blank blank".)
  2. A subtraction machine, where the rightmost number is always smaller than or equal to the left number.
For both problems, clearly describe the proper input and its interpretation, and the expected output and its interpretation, and where does the tape head starts and finishes in relation to the data, and your rule table.

In the past, I gave extra credit to teams that made also a multiplication machine.

26 April
Read Turing's paper on artificial intelligence, "Computing Machinery and Intelligence," available here.

Focus on sections 1, 2, and 6. (Skim sections 3, 4, and 5; they provide a not-very-clear presentation of Turing machines; skim section 7, which considers learning for machines.)

While reading, ask yourself the following (there may be a quiz asking these questions):
  • What is the "imitation game" he describes? How is it set up? What problem is it meant to side-step? (This game is what is now called the "Turing Test.") There are two versions, one with humans, one including a machine. Why two?
  • What are the 9 objections he considers to the claim that computers can be intelligent? (#9 is nuts!)
  • Of the 9 objections, which do you consider strongest? (Don't say #9!) Why?
If you have a minute, you might play with one of the Elizas, like this one.

29 April
Continuing our discussion of Turing's paper on artificial intelligence, "Computing Machinery and Intelligence," available here.

We'll likely have a quiz asking: What are the 9 objections to artificial intelligence that Turing considers. Which is strongest, in your opinion? Why?

We may also consider the Lucas-Penrose argument.

1 May
We'll continue a brief review of the objections that Turing considers, and also our discussion of the Lucas-Penrose argument. Then, on to our next topic: we will introduce the formal definition of Kolmogorov Complexity and Chaitin Randomness.

Your homework: read Chaitin's Introduction to his book, The Limits of Mathematics. This is very good because it reviews much of the material we have discussed in the last half of the semester!

Remember that I'm still taking revisions of Turing Machines.

3 May
Continuing our introduction of Kolmogorov Complexity.

6 May
I've gotten some panicked emails about this, so if you want to take till Wednesday, that's OK.

Team project. You will write and hand in 2 (or 3) Turing-machine and tape specifications. Each machine must do just one thing: print a particular string on the tape. So, your first machine should print 1 below; your second machine should print 2 below; and the third is extra-credit but if you do it your machine must print 3 below. Most importantly, for each of the following strings, create the smallest turing machine and starting-tape-contents that can print the string.
  1. 10101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010101010....(and so on, forever)
  2. 10010101010010111110010000001110110110010010010110100101001000011001001001011101001001000110100100010110101010001001001110110000010101010101011001001000001110100111100000001101111101100100101101001111
  3. 10110111011110111110111111011111110111111110111111111011111111110111111111110111111111111011111111111110111111111111110111111111111111011111111111111110....(and so on, forever)
Remember, number 3 is extra credit.

To really demonstrate Kolmogorov complexity, we should be using a universal turing machine, and writing programs for it; but that would be a bit tricky for us since it would require learning to program some UTM; so we'll just get the flavor of Kolmogorov complexity by using raw Turing machines. We will define the "size" of your program to be the number of rules you use, plus the number of characters in your alphabet, plus the number of characters on your tape before you start. Extra credit to the team with the smallest program, so defined.

The second machine must halt with a tape that shows exactly this string to get full credit; to make things easier for you I'm not asking that the first and last machines halt (we can think of them this way: if we stop them at some time ourselves, it will have the pattern we want in progress of being constructed).

You'll need to hand in a sheet with your rule table, and with your tape's starting contents (you get to decide these, in these problems), and with information about how the tape and machine are at the start, and information about where (for the second problem) the head will be when it is done.

Remember: your task is foremost to get the string on the tape! For example: for problem 2, when your machine is in halt if the answer is on the tape, you've done it. When in doubt, just remember that.

Also, remember that, because some find problem 3 to be hard, it is optional extra-credit. But it's not so hard if you allow yourself an extra symbol, like x, to mark the 1s you've counted already.

In class, we'll look at our last result: the undecidability of randomness.

May 8
I will accept paper drafts up till and including on this day.

May 10
We'll review your homework, and then discuss the undecidability of randomness.

Review of all our major results. Discussion of formal reasoning as akin to empirical work.