Stop Chasing the AI Illusion

The last time I was as delighted by an article in Communications as I was by Peter Naur’s "Turing Lecture: Computing Versus Human Thinking" (Jan. 2007) was May 1979, over "Social Processes and Proofs of Theorems and Programs" by Richard A. De Millo et al. on the formal verification of programs. Maybe I just like to see veterans of the Algol 60 team demolishing the golden calf du jour.

I hope Naur’s article proves to be a sturdy nail in the coffin of "artificial intelligence." Untold millions of U.S. tax dollars have been wasted chasing the AI illusion, which is based on an article of faith (held by atheistic materialists) that the human person is merely a highly evolved biological machine it is theoretically possible to replicate as an electromechanical machine, or computer. Naur argued against the person-as-digital-computer model, employing the principles of the materialist worldview.

Not only has the person-as-computer model diverted energy to the pursuit of AI, it has had a severe negative effect on education in general. If a student is just a fast computer with lots of storage, then education can be viewed simply as the act of storing lots of facts. My children have had to contend with disordered texts written to this model, particularly in mathematics.

It seems that the teaching of concepts and tools with which to derive the specific from the general in order to develop an organized hierarchy has given way to teaching just a bunch of specifics. This model aligns well with the error of "nominalism," denying the ability to know essentials via abstraction while holding that all we can know are particulars.

Returning to Naur’s criticism of philosophy, science, and scholarship, I also presume to offer a correction: Instead of saying "the influence of philosophy on science and scholarship is confusion," one should say the influence of disordered philosophy on science and scholarship is confusion.

Naur distinguished description from investigation of cause, limiting himself to the former, which is crucial for avoiding confusion. Consider the following experiment: I hold a pencil above a table then drop it. The physicist records and analyzes (describes after the fact) what happened, developing theories of motion and gravity that allow us to predict (describe in advance) what will happen when we repeat the experiment. We then test the theories. Confirmed theories allow us to predict reliably and precisely what will happen in similar cases.

We have discovered the "what" and the "how" (description) but not the "why" (cause). Forgetting this distinction, when someone asks why the pencil falls to the table, we mistakenly answer "gravity." In the realm of description (physical science) the experiment is a demonstration and confirmation (proof) of the theories of motion and gravity. But in the realm of causes it is a proof for the existence of God.

Meanwhile, Naur’s assessment of the state of psychotherapy in the "American psychology enterprise" seemed charitable to me. Taking an empirical perspective, he compared it to the state of general medicine in 1800 but passed over two centuries worth of change for the worse in motive and worldview. Naur also aimed much-needed light on the dishonesty, not to say totalitarianism, of the American academic-scientific establishment. Young scholars more interested in truth than money may want to apply elsewhere.

Tim Croy
Houston, TX

Back to Top

To Fight Spam, Try Financial Counterincentives

The article "Spam and the Ongoing Battle for the Inbox" by Joshua Goodman et al. (Feb. 2007) looked for a scientific solution to what is fundamentally a people problem. It came close to proposing a solution in the sidebar "Outbound Spam" but failed to spell it out. Because spam is motivated, perpetuated, and rewarded by financial incentive, it can be stopped only through financial counterincentives.

Consider the following oversimplified solution to make this point. Suppose every email message sent from anywhere in the world is charged an up-front fee of $1. When the message is received, the mail tool instructs the recipient to perform one of three options:

• Open it and, after reading it, label it OK; the sender then receives a credit of $1 in his/her email account;

• Don’t open it or take no action after reading it; the sender receives a credit of 90 cents; or

• Open it and, after reading it, label it spam; the sender receives no credit.

From then on the mail tool filters further email messages from this particular email address. Mail tools do this today but with no consequence to the sender. Email senders willing to spend $1 million to send a million unsolicited email messages would be welcome to do so under this scheme. The revenue thus collected could be put to good use (such as toward R&D for improving life on the Internet).

I’m also aware of the holes in this scheme. For example, someone may point out that constantly labeling email messages as OK from a trusted family member would be tedious. The mail tool would have to include a check box that declares an email address to be safe; from then on, the tool would send the mail-OK message automatically without prompting the recipient.

How could the fees be collected up front? When a user opens an email account with a vendor (even one that provides its service without a subscription fee), the account is classified for a maximum amount, like a credit card.

Few people are able to charge $1 million on their credit cards or send a million email messages at a time. If an email account is good for $100, the sender would be able to send only 100 email messages, then wait for receipt of the electronic credits before sending more (or, alternatively, pay up).

How could such a system be enforced? Again look no further than the credit card. Anything the customer debits but fails to pay is absorbed by the credit card issuer. Such an issuer would still earn a healthy profit, as would the email provider. As with credit cards, the more trustworthy a person’s email habits are over a long period of time, the greater would be his/her mail credit limit. The lesson is: If you want to address a financial problem employ a financial solution.

Sinan Kaptanoglu
Belmont, CA

Joshua Goodman et al. (Feb. 2007) were informative about the efforts being used to block spam, but what’s happening on the other end? What’s being done to shut down the Web sites and email boxes most spammers use, denying them the economic incentive to continue? As for email "stock tips," the only thing I can see to do is to forward them to the U.S. Securities and Exchange Commission and the NASDAQ stock-trading system. A future article on stopping spam would hopefully show people like me (whose email inboxes receive this stuff) what we can do, other than try to ignore it. By the way, if you publish this, please do not include my email address.

Michael N. LeVine
Pacific Grove, CA

Back to Top

Let Everyone Be a Programmer

Douglas Blank’s "Viewpoint" ("Robots Make Computer Science Personal," Dec. 2006) laid out a promising approach to solving a problem that bedevils U.S. industry: not enough programmers. There were plenty in the 1980s when personal computers (such as the Commodore 64 and TRS-80) included an embedded software programming capability based on some version of the BASIC language. [Liberty Basic is available today for Windows PCs (www.libertybasic.com/).] It was easy to learn, and most users quickly began writing programs to support their personal interests or business activities. The original IBM PCs also supported these activities. It’s no wonder that hoards of capable programmers were available just in time for the personal computer revolution.

The turning point seems to have been the introduction of Windows 3.1, which lacked programming capability. The message was: don’t write/just buy programs. At least two requirements must be met in order to reverse this situation: provide programming tools that are easy to learn, and make them widely available. The goal is to allow people to write their own programs.

Blank’s robots represent an interesting approach, but his column didn’t describe the language students will have to use to program them or whether they will be able to program them outside the classroom. Students must be free to explore that language on their own time, not just in class. Moreover, the robots will have to be available to pre-college students, as well. (Each of my three children has a robot, built from kits, none programmable.)

Frederick L. Hills
Herndon, VA

Back to Top

Data Obscurity Via CD

The "News Track" item "A Wikipedia Under Wraps" (Jan. 2007) ignored how data is made available to government intelligence agencies. Such availability is a cause for concern because the data can be changed. However, the solution is easy: Publish the data only on CD, making it accessible only through the CD. No worm would be able to modify the data. Meanwhile, thousands of users would be able to view but not change it. Simple ideas like this are often overlooked.

John A. Carrott
Richmond, VA

Back to Top

Bioinspiration Will Always Cost More Than Nature Pays

Regarding Christof Teuscher’s "Viewpoint" ("Biologically Uninspired Computer Science," Nov. 2006), there are indeed several ways nature is still useful as an inspiration in design.

For example, the way biological neurons work continues to elude us; therefore we cannot replicate or copy it—in the form of silicon or digital electronics—because it’s something we do not fully understand. As a consequence, in the debate between mimicking and replicating we have only one option—mimicking—at least for now. However, we should not be discouraged by neuron behavior, as it most likely matches the extent of our understanding. We can still find useful applications.

While Teuscher’s notion that "Nature isn’t perfect…" may spawn a virtually unlimited range of arguments, nature certainly holds a time advantage over us humans. While we must meet certain economic criteria (even in science), we must remember that funding is based on results, and results must be delivered on a schedule that doesn’t necessarily exist in nature. Nature has all the time it needs, for whatever purpose it has. Not being sure of nature’s purpose, it would perhaps be unwise to argue over whether it is perfect.

I agree with Teuscher that "Good design doesn’t have to resemble nature…" Wheels were developed out of necessity; so were wings that do not flap (though this also reflects our inability to make them flap). Nature did, however, get away without wheels and wings as we know them, solving the sustainable-energy problem, a central issue Teuscher mentioned. We’d like to do the same, but can we?

Nature doesn’t have to pay the price for programmability. Generally, once something is settled, it no longer seems that complex. Nature effectively exploits the underlying physics and chemistry, something we have yet to do and for which we must always pay a price.

We know a great deal about silicon and digital electronics. But are we daring enough to turn our backs on everything we know and start over from scratch? Should we, say, bring biological strategies together with our silicon know-how to shake hands over nanotechnology? For us, price may always be an issue, even if it no longer is for nature.

Bioinspiration may function as a distorting lens, but one should be aware of the fact that lenses usually distort the wide-angle view even as they improve the narrow-angle view. We should look to use bioinspiration whenever we find it in our best interests to do so.

Lucian Prodan
Timisoara, Romania

Back to Top