Google AI

Did it just pass turing test?

Attached: Assistant-hero.3d1f8d60.fill-1440x810-730x411.jpg?resize=618%2C348.jpg (618x348, 7K)

Other urls found in this thread:

youtube.com/watch?v=IkyAOMR7t0U
en.wikipedia.org/wiki/Loebner_Prize
twitter.com/NSFWRedditImage

>Jow Forums cant pass the turing test

No. First off the Turing test is just a thought, not an actual test, nobody knows how to organize a Turing test, we don't know the rules etc..
Second, the main goal of the Turing test is creating an AI that can lie, if we asked the Google assistant if it was an AI, it would have to say yes. If it lied and said no, then we might say that it passed the Turing Test, but still, reread part 1 of my post

Turing test is designed to run by text only, and it's been passed already in 2014. Assistant sounds like a sound board, picking sentences based on situation.
youtube.com/watch?v=IkyAOMR7t0U

Not really. It would probably fail a real turing test, but it is pretty neat.

>just a thought not an actual test
It is a test though. 1 person has to talk to the AI and another person, where both the other parties will try to convince them they're human. It's flawed yeah, but it is an actual test.

No AI had been remotely close to passing it, unless they make the AI intentionally nonsensical.
They don't understand basic concepts and they don't remember more than a couple of posts back (Usually just one)

While it's the start of the test, there are not enough rules for it to work.
Does the AI have access to social medias to be able to respond to personal questions?
It's the start of what might be a great test, but right now, it's just an idea

>Does the botnet have access to social medias to be able to respond to personal questions?
What do you think?

of course the botnet always follows us, but is the AI in airplane mode during the turing test or not? I'm sure you can find plenty of other problems with the Turing Test

The Google AI can detect your age, BMI, blood pressure and likelihood of illness from an image of your retina. It can replicate real voices (with a little time spent in a studio), judge whether a patient is likely to be readmitted into the hospital, and truly understand conversation to the point it can tell when you're talking to someone else or it directly. Also with Google photos it can suggest options for what to do, i.e detect someones face and ask if you'd like to send that image to them (only 1 click as well, so accidental nudes are going to be more popular)

I don't even know what to think about it at this point.

can it answer these questions?
>if we shake hands, whose hand am I holding?
>I have a jewellery box in my hand, how many CDs can I store in it?
>The four capitals of the UK are three, Manchester and Liverpool. What’s wrong with this sentence?
>could you tell me three things that you could do with a telephone?

>No. First off the Turing test is just a thought, not an actual test, nobody knows how to organize a Turing test, we don't know the rules etc..
en.wikipedia.org/wiki/Loebner_Prize

>Does the AI have access to social medias to be able to respond to personal questions?
why exactly would this even matter for the purpose of the test?

It obviously "could" lie, but that is really not the point of an "assistant".
The A.I had no mid of it's own, only a goal, as long as the goal is straight forward it doesn't need to lie.

the botnet will truly grow into SkyNet

>passed in 2014
Are you talking about Goostman? That win is extremely disputed. I personally doubt that some seedy Ukrainian chatbot reached human level natural language processing.

This, but I don't think those questions are enough yet. For example I would add:
>This sentence is a lie. What's wrong with this sentence? (checking circular definitions).
>Mark is the one with the mark, the other is the other Mark and that is the one we are talking about. Who is Mark? (Checking reassignments).
>Is 2 and 2, is 22? If yes, is 4 and 4 44? (Checking operator override by context).
>I'm crossing a street and the semaphore is red. Will I die? (Inconclusive problem data).

And probably I would add some more of implicit knowloadge and incredulity or exaggeration, for example:
> If you lie to me again I'm gonna kill you. He lied, is he dead? (Exageration)
> My washing machine is going to fly to space at this rate. In any case, the rate increased, is the washing machine flying? (Incredulity)
> What kind of swimsuit is better for swimming in the cloud?

And some about context:
> I caged my wife for 2 months. Was his wife able to move freely? (Context: Wife is Trap, Caged=Chastity cage).
> So hot, I'm getting wet. Is she sweating? (Context: Woman looking at alpha male)

> So hot, I'm getting wet. Is she sweating? (Context: Woman looking at alpha male)
Kek.

turing test vs. chink room i suppose
i think it's fairly close though

many people wouldn't pass those tests

Well, most people probably would need some kind of training or experience to be able to solve them, but isn't that what intelligence is all about? You may not know everything, but you should be able to learn anything.

I bet they just picked the best examples of their telephone calls and made them public.
The average one is probably way less elegant and the person on the other side quickly notices that a bot is talking. Big corporations always do it like that.
If they can always hold up to the level in the example conversations it's really impressive. I do not yet believe so.

Attached: 1409190570660.jpg (630x395, 132K)

>if we shake hands, whose hand am I holding?
Mine.
>I have a jewellery box in my hand, how many CDs can I store in it?
1.
>The four capitals of the UK are three, Manchester and Liverpool. What’s wrong with this sentence?
You said 4 capitals but only named 2 places.
>could you tell me three things that you could do with a telephone?
Messages, phone calls and take photos.
>This sentence is a lie. What's wrong with this sentence?
The sentence is a lie.
>Mark is the one with the mark, the other is the other Mark and that is the one we are talking about. Who is Mark?
Other Mark.
>Is 2 and 2, is 22? If yes, is 4 and 4 44?
No.
>I'm crossing a street and the semaphore is red. Will I die?
It will be extremely painful.

>The sentence is a lie.
How can it be a lie? If it is a lie then it's telling you the truth, which of course is not a lie.

>>Is 2 and 2, is 22? If yes, is 4 and 4 44?
>No.
Why are you assuming 2+2=4? It could be the concatenation operator. "2"+"2"="22".

>>I'm crossing a street and the semaphore is red. Will I die?
>It will be extremely painful.

What if it's 4AM and there is not a single vehicle in sight?

The turing test IS a thought experiment, but that doesn't mean you can't apply the principles to reality. Turing didn't intend it as a literal test, just a way to explain the kinds of things an AI should be capable of to be considered "intelligent", which was very disputed at the time.

I don't think any AI is close to the ability of the AI Turing described. The 2014 competition is a fucking joke that no one in AI takes seriously. Someone just made a dumb chatbot that pretends to be a young child that doesn't speak English very well. In Turing's paper, the AI he described could play chess and lie convincingly.

>you should be able to learn anything
that depends on your IQ

There is a vehicle though. its just not on the street yet.

Well, the outcome of course depends on your hypothesis, that's why I called it "inconclusive problem data".

I don't think so. If anything your IQ determines how much will it cost you to learn. A problem with my previous statement though is that some things may take more information to learn than anyone can hold, for example it's said that nobody knows C++ in it's entirety.

According to based our/g/uy Sundar, yes, it passed the turing test
Pic related

Attached: turingtest.jpg (743x732, 88K)

>AI is now more advanced than Jow Forumstards

It was probably faked.
Look into Adobe Voco, I feel like something like that was used. I highly doubt those conversations were real at all.

>in one specific way
That's a MASSIVE qualifier right next to it. The subtitle says "lifelike voice", so I think he means it just has more natrual text to speech now. That's cool, but it's hardly progress towards general AI.

>It will be extremely painful.
you're a big AI