Talk to a Computer that Talks Back

I for one can't wait until computers respond and talk like humans. It'd make navigating those robot customer service phone calls a little more bearable.

Humans? On a customer service telephone line? Really?

A guy can dream

It would be great if TVs could distinguish commercials from programs and mute the sound.

This may be more of an indicator of the changes in what we deep as socially acceptable and its repercussions, in this instance a lowering of the bar set by the Turing test, than it is an indication that computers will begin passing as people.

.

This test was passed by leveraging a niche that was not available until quite recently. While there have undoubtedly always been abrasive adolescent male personalities bound 'to claim he knows anything, but his age makes it perfectly reasonable he doesn't', that hasn't always been something deemed worthy of conversing with or asking opinions of.

.

If we allow lowered social standards to lower the requirements of the Turing Test, I think an argument can be made that the test could have been passed a long time ago. How hard would it have been to get a computer in the 70's to play a convincing game of pinball.... ya know, reminiscent of the mean games played by that deaf dumb blind kid?

Well, how serious can we take this?

One of the panel judges was this guy (Robert Llewellyn):

BS!

https://www.techdirt.com/articles/20140609/07284327524/no-computer-did-not-pass-turing-test-first-time-everyone-should-know-better.shtml

Some days I'm not sure I would pass the Turing test.

That is assuming you can understand a 13 year old boy.

Warwick has repeatedly convinced journalists of breakthroughs in AI when they're nothing of the sort. This chatbot - which does *not* run on a supercomputer - an Arduino or Raspberry Pi would fare just as well - is nothing more than a clever Eliza program. I've written several of these myself; one custom-tailored as a joke which had the dept. secretary convinced that it hated secretarial work and wanted to play games instead. "Look, Pat, we've been working on this stupid document for three days straight. You know as well as I do that Jim isn't going to read it. Let's play Asteroids instead. Wanna?" etc. She actually thought computers were sentient and that hers was "out to get her," and so, by Warwick's standards, it too "passed the Turing Test." This was in 1985.

Warwick's 'news' is not so much about Artificial Intelligence than it is about Natural Stupidity in journalistic circles.

So, they can fool ≥30% of journalists. It's harder to get a cat to follow a string.

Talk to a computer that talks back?

Hello? Haven't these people ever heard of Siri?

Years ago (many, the 1970s roughly), there was a MS DOS GBASIC program that simulated AI as a human remarkably well. I have forgotten what it was called and I don't have the BASIC book anymore where it was printed.....

Maybe someone here can remind me what it was called.....

The program though very simple, made a really good effort and fooled a lot of people at the time....

Did it play the part of a Rogerian psychiatrist?

Example:

>I hate my mother!

You hate your mother?

>Yes. She always favored my brother.

Tell me more about your family.

et cetera.

If so, it was a variant of Eliza of which I've written several (it's not at all hard to do - unless you want it to be very convincing). Eliza programs depend heavily on keyword matching, eg, brother, sister, family, siblings, etc. Emotion words like hate, love. Also reversing the subject/object words in replies: Me-->you, mine-->yours, etc. Tracking context makes it much more convincing but takes a bit of work because some people tend to use pronouns ambiguously: "My father and brother often went hunting together. He brought home a lot of game."

The program needs to know how to recognize when it lacks sufficient information and when to prompt for more: "Who did? Your father or your brother?"

Interestingly, college-age females misused pronouns far more often than males in the same age bracket and some would get exasperated that the program could not resolve the ambiguity. Unfortunately there is no MindReading API.

Some users were reluctant to say much to a machine (but often warmed-up to it after awhile, but not always) If they waited too long to reply, for example, it drew from a large canned repertoire of encouraging statements and questions to get the user to be more chatty. If this failed then it would gradually become more insulting: "Cat got your tongue?" etc., based on the tendency for people to be more likely to speak when they're angry when they might otherwise say nothing. In other words, it goaded them.

The program also maintained a set of indices by 'emotion-word,' for example:

>F--k you!!!

Here the program would increment both the Sex and the Anger indices. These indices steered the program's subsequent dialogue.

All smoke and mirrors.

Oh, one other thing: males tended to brush the program off as just a bit of clever trickery but some females (about 15-20%) would pour their heart out to the bluddy machine, convinced that it was sentient and actually cared about them. Reading the program's notes on these 'patients' was sometimes a real eye-opener.

That was it!! Eliza!!

Many thanks.

I programmed it on my home made NASCOM in 1978 and REALLY got the neighbours up in arms:-

a) A computer that fits in a house! WOW!!

b) It understands me and asks questions!! WOW!!