News

Pro-Palestine Encampment Represents First Major Test for Harvard President Alan Garber

News

Israeli PM Benjamin Netanyahu Condemns Antisemitism at U.S. Colleges Amid Encampment at Harvard

News

‘A Joke’: Nikole Hannah-Jones Says Harvard Should Spend More on Legacy of Slavery Initiative

News

Massachusetts ACLU Demands Harvard Reinstate PSC in Letter

News

LIVE UPDATES: Pro-Palestine Protesters Begin Encampment in Harvard Yard

Focus

Politics and the Turing Test

By B.j. Greenleaf

How do you know that your roommates are conscious? No, not at 4 a.m. after copious libations have reduced them to senseless Neanderthals, but how do you know they are conscious, sentient beings to begin with? You might reason as follows: You are conscious (one can be fairly certain of that, most days) and when you interact with your roommates they behave as if their brains possessed the same characteristics that you attribute to your consciousness. Therefore, you attribute the characteristic of "conscious" to these other entities, your roommates. But this inherently subjective definition of consciousness opens the door wide for beings not based on carbon to be "conscious."

The mathematician Alan Turing proposed a test to determine when computers had achieved this consciousness. Turing's test goes as follows: Imagine you are talking with an unknown entity on, say, instant messenger. If after an unlimited period of time you cannot tell if the entity you are communicating with is a computer or a flesh-and-blood human, then we must treat this computer as conscious. Turing's point was in many ways an epistemological **2) or empirical?** one: We can only define consciousness by the behavior we observe in other entities. So, basically, because you believe your roommates to be conscious only because of the responses that they give in conversation, you must also believe that a computer is conscious if it can fool you into believing it could be a roommate.

So far, no computer has passed an open-ended free-topic Turing test, but I say that we are being too harsh. Just as children progress through the various stages of pre-consciousness to full self-awareness in their toddlerhood, so too computers must slowly approach their sentience. Thus I propose an adjunct to the Turing Test, something that I will immodestly christen the "Greenleaf Test." If a computer can generate sentences that are indistinguishable from political campaign rhetoric, it has passed the Greenleaf test. In other words, if a computer can eventually be indistinguishable from presidential candidates, it has taken its first (baby) step toward full consciousness.

Thus, here in this column I present the first public Greenleaf Test. The following quotations have come from two different sources, one an unnamed presidential candidate (hint: I made the test as easy as I could without lowering the bar below the two party system), and the other a simple "babbler" computer program (CS-51 students might recognize this program as the Markov-based babbler of assignment 8). I allowed the program to use some general text from the congressional record so that it might "learn" the fine nuances of political rhetoric. And here, in no particular order, are the quotations (four quotes come from each source):

1) "Once their initial term of service has been honorably served, a veteran is eligible to receive the basic monthly education benefit for our nation's seniors and for our nation's children, parents and teachers by sending more dollars directly to the classroom."

2) "I think it's important for those of us in a position of responsibility to be firm in sharing our experiences, to understand that the babies out of wedlock is a very difficult chore for mom and baby alike."

3) "What I am against is quotas. I am against hard quotas, quotas they basically delineate based upon whatever. However they delineate, quotas, I think vulcanize society."

4) "All credible research has shown that prevention and early intervention initiatives, combined with a continuum of services aimed at high-risk, cutting edge, fundamental, and may not be used for specialized courses, such as women and youth, increase prevention and treatment activities for sexually transmitted disease and substance abuse which contribute greatly to this crisis, expand research, increase Medicaid funding, bring our programs to smaller cities and rural areas and greatly increase the technical assistance that will enable our community-based organizations to take advantage of the many excellent technology-related courses sponsored by companies like Microsoft or Novell."

5) "We want our teachers to be trained so they can meet the obligations, their obligations as teachers. We want them to know how to teach the science of reading. In order to make sure there's not this kind of federal cufflink."

6) "Let us dispense with the rhetoric and look at the difference, the difference between foreign substances and the patient's own cells."

7) "NIH is very important to that institution providing money to back basic research."

8) "And, you know, hopefully, condoms will work, but it hasn't worked."

In my humble opinion, the machine (or maybe I should say the distinguished representative of a digital constituency) has passed the test with flying colors, and, as Turing's logic inescapably leads us to conclude, has earned the respect accorded any high-ranking political hopeful. In fact, I have been looking into getting it into the presidential debates but with no luck as yet. We have witnessed a serious step in the evolution of computer consciousness. While we cannot really think of a computer running this "babbler" program as having a fully human-like intelligence, we can consider it to be a viable presidential candidate. Well, to be fair, this "artificial" candidate has one major advantage over its flesh-and-blood cousin--with a flick of a switch its inane babbling is silenced.

B.J. Greenleaf '01 is a physics concentrator in Mather House. His column appears on alternate Tuesdays.

Want to keep up with breaking news? Subscribe to our email newsletter.

Tags
Focus