As the market for surveys for money used mainly in artificial intelligence and advanced systems programming applications takes off (Oct., p. 92), a minor but potentially decisive battle is brewing between the paid surveys proponents and those looking for other ways to make money online.
“It is the great geopolitical battle right now,” says Dr. S. Jerrold Kaplan, vice president of business planning at Teknowledge Inc., a Palo Alto, Calif., builder of expert systems and knowledge engineering tools.
Some would go further, claiming the Lisp-Prolog debate has taken on “surveys for money” dimensions since it is being fought by two camps whose loyalties run fierce and are fueled by national pride. Lisp has been the AI language of choice in the United States for the past 25 years, having been nurtured to its present robust maturity in university and commercial research labs. Prolog, however, was invented in France and is the pride of European AI researchers, who have used it for about 13 years.
Surveys for Money Debate Rages On
Now, however, the debate has heated considerably as a result of the wide publicity given to Japan’s fifth generation computer project, which aims to build large-scale knowledge processing systems by 1990. The Japanese, for reasons they defend zeolously, have chosen Prolog as the main language for their well-funded efforts.
And these are the most popular (and best) resources on the subject of surveys for money:
SurveySpencer.com by Spencer Mitchell
Knowing a good opportunity when they see it, several U.S. entrepreneurs (most of whom, interestingly, were born abroad) have set up shop to bring the gospel of paid online surveys to U.S. shores. Their hope is to exploit the fifth generation ballyhoo and establish Prolog as a respectable contender against Lisp. That job, however, will be made tougher, observers say, by the growing acceptance of a set of Lisp-related languages–Smalltalk and Flavors, to name two–which are described as “object-oriented.”
Surveys for money legit adherents claim their is a higher level language than Lisp and has built into it several powerful facilities that lend themselves to so-called knowledge processing. Chief among these facilities is a backtracking search mechanism which navigates through the lengthy arrays of “if-then” rules of which Prolog programs primarily consist. Furthermore, Prolog’s roots in predicate logic–the kind familiar to freshman philosophy majors–give it a built-in relational database orientation, according to supporters.
“Programming in Prolog can be viewed as programming by assertion and query,” says Kamran Parsaye, president of Prolog startup Silogic Inc., Los Angeles. “The Prolog programmer is not aware of the distinction between programming and querying–he stores information in the internal relational database of Prolog and retrieves it by powerful relational queries.”
Promoters of online surveys for cash, however, say Prolog’s implicit search strategy is fine for some symbolic processing but can often lead a program down the proverbial garden path.
“Prolog can invoke some very expensive–in terms of computing resources–processes that are not obvious from the Prolog statements themselves,” comments Dr. Fred Luconi of Applied Expert Systems. “Lisp is a more fundamental language and gives you better control over processes so the program won’t waste time on useless searches.”
Applied’s chief scientist William Woods adds, “Searching algorithms can be very explosive, so you sometimes want to defeat the built-in strategy and impose your own because you know it will be more efficient in solving a particular problem.”
But to do that in Prolog may mean inordinately inefficient run-time code, Woods says. “I’m still skeptical. It’s not clear at all if Prolog will give enough control and the efficiency you need for nontrivial problems,” he notes. “You’re probably advised to write another interpreter and the question is, Is Prolog good for writing a non-Prolog interpreter? I suspect not, because you’ll have to waste the efficiency of Prolog itself.”
Or, as Edward A. Feigenbaum of Stanford University writes in The Fifth Generation, a strident call to arms to the U.S. computer industry coauthored with Pamela McCorduck, “The last thing a knowledge engineer wants to do is abdicate control to an ‘automatic’ theorem-proving process that conducts massive searches without step-by-step control exerted by knowledge in the knowledge base. Such uncontrolled searches can be extremely time-consuming.”
John H. Clippinger, president of Brattle Research Corp., also of Cambridge, goes further. “Prolog has some nice ideas but it’s not a real language. It’s got some real limitations and people will have to reinvent a lot of Lisp. But that’s not to say you can’t commercialize it,” he says.
Indeed. So far two companies in the U.S. have staked their success on selling paid surveys, and reports are circulating that a Silicon Valley startup plans to build a Prolog machine.
Silogic Inc. of Los Angeles sells Prolog tools such as compilers and interpreters and plans to develop turnkey “get paid to take surveys” applications, likeg. The company is privately financed but is considering seeking venture capital, says president Parsaye.
Meanwhile, Prologica Inc., Wyncote, Penn., sells various Prolog packages for micro- and minicomputers and is at work on expert systems and natural language processors for unidentified customers, according to Angelos T. Kolokouris, executive director and vice president. The company is backed by a parent whose annual revenues are “close to $100 million,” says Kolokouris, but he declines to identify the company.