In another writeup some time ago, I explored the question "Is Everything artificially intelligent?" by using a modified Turing test in the Search box. Visit that node to see the questionable results.

Recently, I decided to revisit my technique for carrying out the conversation. My new rules were as follows, with modifications in bold:

  1. All questions are entered into the search box in hopes of finding a match.
  2. When a perfect match isn't found (which was always), the first node in the list of possible matches is selected.
  3. If the node is actually a nodeshell, the next node in the list of possible matches is used. This is repeated until a non-nodeshell is encountered.
  4. The first complete sentence (no fragments) of the highest-rated writeup in that node is considered Everything's answer. A sentence ends with a period, question mark, or exclamation mark, or else is marked by the end of a paragraph. If a complete sentence isn't found, the next search result is used.
This modification required the expenditure of some votes to determine which was rated the highest, but this time around I had them to spare. The first conversation is as follows:
Hi, how are you today?
- There are many possible ways, and people seem to forget the varieties.

Well, um, what way are you right now?
- I seem to be completely crap at my chosen vocation.

Oh, I'm sorry. What vocation is that?
- I thought I was a fan of it, really.

I take it you'd rather not share?
- To borrow (ok steal ouright) from Singles, "Your line is that you don't have a line."

Do you think I'm trying to hit on you or something?
- he says this while the three siblings are trying to sell some sort of scout cookies to Albert Einstein (incidentally one of the best episodes, EVER!)

Are you talking about me, or someone else?
- It could take years, and whole parades of other words could pass through this room before he would be through with the word "love."

Ah, so you're having romantic difficulties.
- I am comfortable around you.

But not around this other person, I take it?
- You know, I've said this much before, but jerk boys can be very fun crushes.

I'll take that as a yes. Does this boy have a name?
- Its also fun to say.

But what is it?
- Time and time again, we're shown that indeed, it's not what you know, but who you know.

Are you saying it's someone famous?
- several days later i do a user search on their name and find that they have only three nodes-- their home node, a nodeshell, and a mediocre writeup with no links at all.

(Clearly not interested in sharing, so I continue) No common interests, then. Bummer.
- There is no conflict of interest.

Nono, I said COMMON interests.
- "Common sense is nothing more than a deposit of prejudices laid down in the mind before you reach eighteen."

Are you misunderstanding, or are you insulting me?
- I recently discovered the origin of this phrase.

(Giving up) And what might that be?
- I paint a simple picture: Me and my buddy Evan in a raquetball court.

As you can see, the conversation still dead-ended before very long, but it did have potential there for a while and was certainly much more coherent than under the previous rules.

In order to try and keep the "train of thought" a little more consistent, I then tried asking questions by searching for my question and the previous answer. The results, unfortunately, weren't any better.

The main problem with using this type of Turing Test on E2's search engine is that E2 doesn't have any concept of context. Furthermore, it's a bit schitzoid in that it cannot keep track of the conversational topic. The real pitfall with assuming that E2 is conversationally intelligent is that, conversationally speaking, it has the interactive capacity of a two-year-old, moving from topic to topic without any coherent flow. It's too constantly distracted by the new subjects that come up in the search engine to follow its own "train of thought", which ends up being more like a "train of thought wreck."

The best solution for this kind of problem, I think, would be for E2 to eventually have the capacity to perform boolean searches that extend into the bodies of write-ups. A node title alone may not have anything to do with the content that fills a nodeshell (as many of us know all too well). If E2 could parse or extrapolate a write-up's content and compare it the context of a searched phrase, E2's conversational accuracy would be more realistic- even comprehensible. Right now, though, it's like having a conversation with a young Robin Williams who is suffering from ADHD and is on crack cocaine.

Doing a Turing Test on E2 now, as it stands, will have peculiar results across the board, no matter how many different angles of attack you use. The final analysis from a psychological point of view will be woefully bleak, making it seem like E2 is either inherently stubborn, bi-polar, schitzophrenic, depressed, mildly psychotic and antisocial in the extreme or that it's just plain bonkers. In its current incarnation, E2 would be sent to a mental clinic if it was to be interviewed by a hardline Humanist psychologist. I'd bet even money on it.

And let's not even get into the various bots that work within E2's matrix, like the EDB, Cool Man Eddie and Klaproth. A shrink's take on those would be that E2 is suffering from MPD (multiple personality disorder) and would prescribe about a metric ton of thorazine for the patient.

Log in or register to write something here or to contact authors.