Three farewells

We bid farewell and requiescant in pace to three people this week:

Jean DeLoach, a long-time family friend in Valdosta, one of my mother's close friends at the time I was born;

Patsy Smith, legendary Valdosta art teacher, whom I had for sixth grade art and again briefly in high school (at VHS) before I joined the debate team and had to change my schedule;

Charles Hudson, author of The Southeastern Indians, whose anthropology course I very much enjoyed as a UGA undergraduate.

Permanent link to this entry

What are tensors?

Mathematics day in the Daily Notebook...

The latest wave of AI algorithms "use tensors," which has led me to object to the ambiguous way mathematicians use the word "tensor." I've seen the word used to mean 3 different things:

(1) An n-dimensional array of numbers, where n (the "rank") is 0 (scalar), 1 (vector), 2 (matrix), or higher.

(2) Like (1), but implying that n > 2, or else it would be called a matrix, a vector, or a scalar.

(3) A tensor as defined in (1) understood as a transformation applied to a tensor of lower rank. Thus a matrix is a transformation applied to vectors; a tensor of rank 3 can transform matrices; and so on.

I think (3) may have been the origin of the term. The quintessential example is the stress tensor in physics. The stress on a solid object is a transformation of 3-element vectors. If you are at any point in an object under stress, you can look in any direction (a 3-element vector) and there is a stress applicable to the direction you're looking; you can think of it as distorting the vector. The stress tensor is in fact a matrix, a tensor in sense (1) but not (2).

I know (1), (2), and (3) are closely related, but they still reflect different ways of understanding tensors from the human point of view. It's annoying to get ready to sink my mental teeth into (2) and then find someone is merely doing (3) with a matrix, using familiar linear algebra.

Permanent link to this entry

New hat

I am literally wearing a new hat these days; the consulting business is going well. (Photo by Melody.)

So far only one had has been made; there may be a few more, but I'm not planning a mass giveaway. Vistaprint made it, quite affordably, from a bitmap image. There may also be a Covington Innovations coffee cup in the future.

Permanent link to this entry

Do we really want a totally free market for labor?

I recently expressed agreement with this court decision, which ruled that many unpaid internships are a violation of minimum-wage laws. During the tight labor market of recent years, some companies have been availing themselves of "interns" as a source of free labor, not just recipients of education. The court wants to restrict that.

I was surprised to get several replies from people who think we shouldn't have minimum-wage laws (nor, presumably, other wage-and-hour laws or industrial safety laws). Sink or swim, they say. Setting a minimum wage just causes unemployment.

Well... I know a minimum wage causes a certain amount of unemployment. If the minimum wage is $6 per hour and you can't bring somebody $6 per hour of profit, you can't have a job. The thing is, without a minimum wage, you wouldn't have a job you could live on, either, and neither would other people. The minimum wage forces employers to manage labor in such a way that it is reasonably profitable. It prevents a tradition from arising of paying people far less than a minimum wage and making them work in degrading conditions.

But won't the free market take care of that? If a job isn't tolerable, won't people just refuse to take it?

I reply: Those who cannot remember the past are doomed to repeat it, and I'm afraid the current vogue for libertarianism, combined with ignorance of what has historically gone wrong, may take us back to the bad old days. (Like the photo by Lewis Hine above, of child laborers in 1908.)

Free markets are a great way to allocate resources, but they're not perfect. They fail when power is very unevenly distributed. For example, in The Grapes of Wrath we read a realistic account of people who gave up farming during the Great Depression to try to be migrant farm workers. Employers could easily make misleading offers, and the workers couldn't communicate among each other to find out how bad the situation really was. The free market failed to work as intended. Labor unions were one early response to this kind of market failures. Modern means of communication are another.

Anybody who wants to apply Christian values to economics also needs to know about Rerum Novarum, the famous encyclical in which Pope Leo XIII, assisted by top-notch Catholic economists, denounced both Communism and a naive form of capitalism. His point: Saying "we have a free market" or "they're doing everything voluntarily" doesn't exempt you from responsibility for the way you treat people. Workers have responsibilities to do an honest day's work, refrain from rioting and vandalism, etc.; employers have a duty to pay a living wage, have reasonable work expectations, and generally be aware of what they're doing to people. Even if you disagree with the details (and if you're not Catholic, Rerum Novarum certainly does not bind you as authoritative), this is something you need to know about. It is not left-wing drivel. An important sequel to it was written by John Paul II, the man who brought down Communism.

Finally: If you claim to be a Christian but get your moral philosophy from Ayn Rand, get out of my face. Rand is one of the most vehement enemies Christendom ever had, and it's bizarre that so many of my fellow Christians have been bamboozled into thinking that Christianity equals extreme capitalism equals Rand.

Permanent link to this entry

NAACL-2013

On Monday (June 10) I attended one day of the North American meeting of the Association for Computational Linguistics (NAACL-2013; proceedings here and here.

I hadn't been to an ACL meeting since about 1998. I had a delightful time reconnecting with colleagues. There was a lot of talk about the "Covington parser," which is actually (I think) one of Joakim Nivre's dependency parsers based on a paper of mine from years ago. I'm better known than I thought. Also vindicated: in the late 1980s everybody was telling me dependency parsing was worthless, and nowadays everybody is doing it.

The meeting was in the Westin Hotel in Atlanta. As you see in the picture, the poster session was held at 200 Peachtree, formerly the Davison's department store in downtown Atlanta, where Melody was occasionally taken shopping as a child. Also, while in the neighborhood I located the place where I bought my Olympus OM-1 camera in 1976. Foto-Crafts-Atlanta was at 135 Carnegie Way, which is now a Courtyard by Marriott hotel. That's Atlanta: you identify a place by what is there now, what was there ten years earlier, what else was there ten years before that...

Permanent link to this entry

Against cloud computing?

I don't quite trust the "cloud computing" vogue, and Richard M. Stallman, free-software radical, has now expressed a position like my own, but stronger.

Key idea: Cloud computing threatens to make us dependent on just a few, expensive, centrally controlled services, which is the opposite of what microcomputing and the Internet have been doing for us until now.

Back around 1972, just before microprocessors were invented, a lot of us imagined a future worldwide mainframe computing service, with terminals in every home and office. Naturally, we expected it to be centrally controlled. (Control Data's CYBERNET scientific computing service foreshadowed this and was really useful to its subscribers.)

Microchips came along and changed all that. By 1982, you could have a reasonably powerful computer that ran all by itself, without dependence on a mainframe service provider.

Meanwhile, the Defense Department's high-reliability network technology became the Internet, crucially with no central control point — it is basically a peer-to-peer network throughout, with all control and relaying functions shared among computers owned by anybody and everybody. That's right — nobody controls the Internet. That's what made it the most powerful force for freedom since the printing press.

Enter, now, cloud computing. Instead of letting your computer run by itself and store its own data, you're supposed to rely constantly on one or more central service providers. While this is useful in some ways, it also threatens to take us back to the bad old days, when there were central control points, central points of failure, and central companies to whom money had to be paid.

Fifteen years ago, web hosting and e-mail were handled by thousands of different independent ISPs. Today I am starting to meet people who think all e-mail should be on Gmail.com and all files should be on Google Docs or Dropbox... in short, they want just a few near-monopolies instead of the radically decentralized network we used to have.

This is not good. Perhaps the strongest force toward centralization comes from search engines such as Google and Bing. The world can only support a few of them because of economies of scale (they benefit greatly from being huge), yet everybody needs them. Could we reinvent search engines to work more like the Domain Name System, based on sharing information between vast numbers of voluntarily participating machines? Just a thought... maybe a way to reduce the pressure toward central control.

My own approach to cloud computing is mixed. I have my own e-mail and web domain (covingtoninnovations.com), hosted by an ISP in Michigan (and formerly hosted somewhere else; I can move it whenever I want to). I use Dropbox for a few unimportant things, but mostly I do cloud computing on my own server so that I can sync my laptops, from anywhere, to a computer in my own house.

Let's reverse the trend of turning computing into cable TV, which seems to be the direction in which we're headed. Remember when you could get TV without paying for a cable subscription? No? It was done with antennas...

Permanent link to this entry

Data mining — a term being hijacked?

I fear that the news media are hijacking the name of one of the things I do professionally.

They are using the phrase "data mining" to mean "unauthorized snooping."

That's not what data mining is. Data mining is the use of computers to search for patterns in data. It's very valuable. Nobody objects when a grocery store data-mines its own sales data to find out which products are often bought together, or when a telephone company data-mines its traffic data to figure out where to build more lines.

(Click on the picture for a good book about data mining.)

When telephone traffic records go to the federal government, the controversial thing isn't data mining — it's who's getting whose data.

Please, let's not use the term "data mining" to mean unauthorized disclosure of data to a third party.

Permanent link to this entry

Telephone follies

For decades I have answered the telephone with the words, "This is Michael Covington, may I help you?"

Increasingly, callers are unable to understand a word of that. It is as if I had said, "Εδώ είναι ο Michael Covington. Πώς μπορώ να σας βοηθήσω?"

All they want is "Hello." They don't understand anything else.

The other things they don't understand are, "Who is calling?" and "To whom do you wish to speak?"

(I know using "whom" is like speaking Latin these days, but I persist in it. I figure that if you call me, you ought to understand my language.)

Lately I've had a couple of calls for "Who dat?" and one who only said, "Don't call me 'bout no break-in!" Neither of them had a clue who I was.

Permanent link to this entry

When ink spreads out on the nib of a fountain pen

Some of my fountain pens have a tendency to get ink all over the metal surface of the nib. This could be caused by rubbing the nib against the cap as you take the cap off, but I checked carefully to make sure that wasn't happening. On most pens, it is not actually a problem; the ink doesn't get on the cap or your fingers, unless it's a Sheaffer "inlaid" nib, which doesn't have a grip separate from the nib itself. Anyhow it looks messy.

After investigating the problem at some length, I've found that it's much worse with some pens than others; some nibs may have a very light coating of wax (which might need restoring) to make the surface repel watery solutions; but mainly, it's the ink.

In particular, it's the amount of wetting agent (detergent) in the ink. You may sometimes notice that an ink has a tendency to foam if the bottle is shaken before you take the lid off, or if a pen sends bubbles of air through it. That's an indication of lots of detergent.

I have the worst problem with a relatively new Hunter ink; the least, with Sheaffer blue and Waterman blue; several reds or purples are in between.

So, among other things, the bottom line is, with Sheaffer inlaid nibs, use Sheaffer ink. And with everything else, keep your eyes open!

Permanent link to this entry

Point of theology

Young-earth creationists often assert that God created the earth recently but with the appearance of age, which includes false evidence of numerous past events that didn't happen, from supernovae to the life and death of particular animals.

A fellow Christian whose judgement I respect points out that this distorts the character of God much more than a mistake about morality or ecclesiology would. If God is a deliberate deceiver, what does that tell us about His integrity and about the nature of revelation? You've almost moved out of monotheism as we know it. You'd be worshipping a trickster god like Loki whom you can't trust at all.

This is perhaps the biggest reason I'm not a young-earth creationist. If someone were to show that the earth is, after all, young, they would have to account for all the evidence that it apparently is old. And by "account" I do not mean "just say God tricked us."

Permanent link to this entry

Short economic notes

Has anyone noticed that interest rates are starting to go up? Not the prime rate, but various bond rates and loan rates.

Back in the 1970s we were being told that the "population explosion" would lead to mass starvation within our lifetimes. Actually, world poverty is being conquered, and it's due to economic development. Whether or not you accept all the explanations offered in that particular article, the fact remains that poverty is diminishing.

This leads to three more remarks: (1) The United States is no longer the only big prosperous nation, and the people who still think it ought to be need a reality check. (2) One reason gasoline is expensive is that more people around the world want to buy it. (3) Pollution and carbon footprint are largely developing-nation problems, not North American and western European problems. China and India are a lot more populous than the U.S. and will soon be much larger energy consumers, if they aren't already.

Permanent link to this entry

An affordable laptop for college

A friend asked me how to find an affordable laptop computer for his daughter to take to college, and that led me to jot down a few notes here. I'm sure not everybody will agree with everything, but here's how I see it. This advice is for students who are not computer experts and do not have specialized needs.

You don't have to buy one machine for four years. For a student whose computer needs are unclear or uncertain, consider buying the cheapest usable machine, maybe secondhand or handed down from another family member, and making another purchase one or two years later. If you try to buy a computer that will be up-to-date four years from now, it will be overpriced today. Don't buy too much before you know what you need.

PC, not Macintosh. Let's face it, Macs are like BMWs — well-built but expensive — and you're looking for a Chevy. Besides, Macs don't run PC software. Don't assume that you have to have a Mac just because you're doing art or music or because the college bookstore pushes Macs (and makes a big profit on them). And if you've been frightened away from PCs ("only a Mac is easy to use"), overcome your fear and get back to reality. If Windows were unusable, 95% of the world's computer users wouldn't be using it. And you don't want to come out of college unable to use it, do you?

My experience is that Asus laptops are very similar to Apple Macbooks in construction and quality, but they only cost half as much because they run Windows. I also like Lenovo and Dell. Don't mix up Asus with Acer, which is low-priced and, in my limited experience, not as solidly built as the competitors.

You don't need high-end graphics. The difference between a $400 laptop and a $700 laptop is likely to be an ATI or NVIDIA graphics system, needed for interactive game playing but not, in my experience, needed for anything else, not even photo and video editing. Similarly, you don't need an enormous disk drive.

You don't have to buy Microsoft Office. If you do buy it, get the student edition. In fact, look for student editions of other expensive software, such as Adobe products, through your college bookstore or JourneyEd. The discounts are often enormous. But you can use free software such as OpenOffice for practically all of your work.

Be skeptical of the college bookstore. As I said, student prices on software are often excellent. But "student discounts" on computers themselves (laptop or desktop) probably won't undercut Wal-Mart, Office Depot, or Micro Center. Shop where working people shop. Find out what kind of laptops the faculty of the college are using.

Finally: I have mixed feelings about used laptops. Laptops tend to get beaten up and worn out within just a few years. Having said that, a good used laptop with at least Windows Vista on it can be a bargain. Batteries need replacing after a few years; keyboards and disk drives are easy to replace too, and I have routinely done this on older laptops that I'm using myself.

Permanent link to this entry

It's a book

Our latest book is the eleventh edition of this classic, and now it has five authors. Click on the picture for more information.

Permanent link to this entry