Art,
True enough, with one caveat regarding human behavior and math. We as humans do complex math on our heads all the time. Consider the following scenario;

User: Can we make it to Portland or do we need to stop for gas?
Bot: We should stop, there’s a filling station at the next exit

flowcharted (I still use them I know they are supposed to have died off long ago) LOL

Bot access GPS for automobiles current location
Bot accesses map service for distance to Portland from current location
Bot accesses sensor for auto speed
Bot calculates rate of consumption at speed averaged over current trip
Bot accesses current fuel level
Bot accesses previous trips, or uses Bayesian probability logic to predict likelihood that user will stop for food or rest stop
Bot calculates increased fuel consumption based on restarting engine
Bot determines probability for making trip without stopping

And that’s simplified. If someone asks you to solve an algebraic equation, you might balk yet you do it thousands of times a day without conscious thought in order to have an “average” conversation.

Really fairly mind boggling when you get into it

Vince

I like your questions, Art. A Turing Test in the spirit of Turing’s argument should ask questions that test intelligence, and judge answers by how well they are answered, instead of testing for human character. Turing himself exampled questions about math, chess and poetry, not coincidentally because these were all considered proof of intelligence at the time.
I would exclude the rhyme and letter games though. As those have very little practical use elsewhere, it is much more tempting to hardcode specific systems for these games than it is worth the trouble of raising general AI for. I particularly like the word disambiguation questions because I’ve done a lot of work on that ability.

Vincent, I believe there have been some covert Turing Tests (with unaware judges) already. If not in the classic Turing Test format, then certainly in video games and unintentionally in real-life telephone operating centres. Not informing the judge that there is an AI raises the benefit of doubt % to the max and undermines the purpose of interrogation for the “human” aspect. However, I do feel that rating the helpfulness of a program alongside its ability to get along with humans smoothly form a test worth passing.

Vincent,

The examples you’ve provided work well in / with my smart phone, especially if I employ the Google microphone on my Android. I have been in the midst of Washington D.C. traffic at the end of the day approaching “rush hour” traffic and with a tap on the Mic and a, “Take me home” command, the rest is easy following the on-screen and verbal guidance.

What we’re discussing here, however is the methodology of a chatbot, not smart phone app, passing the Turing Test.

What if the judges were NOT informed how many “units” would be people and how many would be computer programs?
What if they were told there was a 50/50 ratio of humans and computers and they need to decide which is which?

Lastly, The math tests, need to be kept simple as most people (average / ordinary) do not use much beyond basic math.
Some have difficulty figuring miles per gallon or Distance = Rate X Time equations. IN order to keep things on a more level playing field, the Test criteria should bring things down to earth a bit. e.g. - the bot should not know any more than a human should know.

Just some thoughts….

If you don’t tell the judges how many programs there are, I think it gets more interesting because it removes the lame 50/50 guess. Meaning you’d start with 100% odds of being mistaken for a human instead of 50% odds otherwise.
Simultaneous conversations would then no longer have purpose since one might also be talking to two humans or two programs simultaneously, so the judge should have to talk to one entity at a time, and compare the responses to what he imagines a human would say. The purpose of interrogation would still stand, and even receive more focus, while the benefit of doubt would also be increased.
It would be well in accordance to the argument that we assume other humans are intelligent out of courtesy, and why not grant the same courtesy to beings that act human (This argument equals ignorance to proof, but it’s one of the ideas behind the test).

Didn’t one of the 90’s Loebner Prizes have a contest where only 4 of 12 were programs, and the judges weren’t told? It would be interesting to read a report.

Miles? Gallons? What are these un-European units you speak of?

Absolutely terrific with the exception of our Miles and gallons.
#######
The late Robert Lee Frost was an American poet. His work was initially published in England before it was published in America.
#######
With that in mind how awful would his poem have read if he’d said, “...and kilometers to go before I sleep.”?

Regarding the gallon:
The imperial (UK) gallon, now defined as exactly 4.54609 litres.
The US gallon, which is equal to approximately 3.785 litres.
difference of maybe 32 ounces…go ahead…ask about them?

It matters not how you measure, as long as you keep in mind that a string is twice as long as half it’s length!!:)