I agree that making chatbots passing the Turing Test will probably be something so common in the future that it will be a student project of medium complexity.

However, we are not there yet. If someone had this today, how valuable would this be to Apple, Google, Facebook, Microsoft and Amazon?

As for the definition of what the Turing Test is, it's definitely a fuzzy subject. My own arbitrary definition is "ability to convince a human that he's talking to another human after a sustained (based on time or length of) conversation whereas human is aware that there's a possibility that his interlocutor maybe a machine".

So, it's more of convincing a judge in Loebner Prize competition than a random troll on Twitter.

Woah! I thought they pretty much already have passed it? Remember Ashley Madison? You had ~12 million heterosexual men (that were cheaters, 6 million 'active' users) trying to talk to ~12k heterosexual women (also cheaters, 10k 'active' accounts). It ends up being about a 1:13,000 ratio. Not only that but MANY of these men had paid actual real money to the site in order to do so, and then continued to do so. The only real conclusion was that most of the men were talking to bots that the site had made up.

Ok, lets get this straight: ~6 million real human men paid real money that they earned through their labor or whatever to talk to bots and then paid more real money to do it again. Admittedly, they are 'cheaters', but 6 million men must have an IQ distribution nearly identical to that of the general population, i.e. they represent heterosexual human males in general. And yes, they were trying to get laid, these conversations are likely pretty brief, and mammalian males are not generally known for using their neocortex during mating.

Still, I think that 'counts' as far as passing the Turing Test. Yes, now we can move the goal posts to say that the bot has to teach me something, or guess what I was thinking, or generally be better than a man on tinder. But as a first pass of the TT, I think we have been here for a few years now.

https://en.wikipedia.org/wiki/Ashley_Madison_data_breach

There's no reason 6 million non-randomly selected men would be likely to have an IQ distribution similar to the general population. You can't make up for non-random selection with a larger sample size.

Ok fine, but then how far off of the mean should they then be? They aren't all super smart nor are mentally deficient as they have to be able to function in society to make money enough to pay for the service. At most this is what, an IQ mean of 75-125. So that then means at the low end, TT chat-bots can fool human males of IQs of 75. That's pretty darn good and that was 4+ years ago.

One application I predict is spam, from nornal spam being less detectable to those spam bots that add you on chat platforms and try to get your card info, responding to what you type.

Turing test is relatively easy to pass if you creatively design your program to roleplay an irrational character. See this program https://en.wikipedia.org/wiki/Eugene_Goostman

Obviously such programs, while being works of art, are not interesting from AI/Machine Learning point of view.

We've already had a few chatbots pass a turing test. Just none that are particularly sophisticated. One in particular just acted "cheeky" to throw off the other person.