ProfessorFalken
Registered user
Registered: 12-2017
|
Reply | Quote
|
|
Re: Pretty sure Google's new talking AI just beat the Turing test
heres the vid - https://www.youtube.com/watch?v=lXUQ-DdSDoE
jesus christ, if google arent bullshitting - and I suspect theyve made 100s of calls and these are just the best results - even so im still staggered - amazing!
|
5/9/2018, 10:04 am
|
Link to this post
PM ProfessorFalken
Blog
|
greendocnowciv
Registered: 11-2017
|
Reply | Quote
|
|
Re: Pretty sure Google's new talking AI just beat the Turing test
Spike, Prof Falken,
Spike,
Thanks, good to know we are that far.
Prof Falken,
Thanks to you, too. I wasn't going to pursue watching this, but now that I have I am glad I did.
The two examples that the Google CEO played on the big screen for his audience were definite "Turing Test Mark I" passing grade examples.
Two retail outlets, a hairdresser and a restruant appointment call both passing the test.
Next, most of us can imagine it passing a test with a person more attentive to nuances of detail, more than the often glitchy customer service call where the business is used to glitch people.
In this vid example, the computer was much better than many people I have dealt with. And sadly, better than I have been more than once in my life.
Seeing this vid, I am more convinced than ever that our time - not some far future - will see Mr Data, Westworld, and AMC's HUMANS.
The progression curve to there is clear. For those who doubted it - watch that vid, and see it progressing as we can be sure it will.
|
5/9/2018, 11:13 am
|
Link to this post
PM greendocnowciv
Blog
|
Extropia DaSilva
Moderator
Registered: 12-2017
|
Reply | Quote
|
|
Re: Pretty sure Google's new talking AI just beat the Turing test
Sounds more like a variant of the Feigenbaum test to me. For those who do not know, the Feigenbaum test has an AI compete against professionals in some particular domain (neurosurgeons, say) and the conversation is limited to on-topic subjects. Thus, the Feigenbaum test is more restricted than a Turing test, which can broach any topic whatsoever.
Is reserving a seat a conversational challenge that is less or more limited than being interviewed by top experts in a particular domain about that field of expertise? I would say it is more limited. So at best what we have here is a program passing an easy peasy Feigenbaum test. The Turing test remains beyond the capabilities of software. For now.
|
5/9/2018, 1:55 pm
|
Link to this post
PM Extropia DaSilva
Blog
|
andmar74
Registered user
Registered: 12-2017
|
Reply | Quote
|
|
Re: Pretty sure Google's new talking AI just beat the Turing test
Impressive but far from passing a Turing test.
I wonder, why the pause before the AI answers? That's a give away.
|
5/9/2018, 11:24 pm
|
Link to this post
PM andmar74
Blog
|