The full conversation here: https://imgur.com/a/5wjVFiS
Now I understand that chat support people copy and paste messages, that there’s a very Google voice that they have to use, that they are trained on timing. I get that the "I see.", "Let me check" are standard Google. Maybe the "Are you still there" is just automatically injected into the chat after inactivity no worries there.
Yet I am convinced I just talked to a fully functional AI.
At first the similarity I was seeing just came from the Duplex video at the IO and the time I’ve spent with Google Assistant. But as the conversation progressed I just could not let go of the notion that this was a Turing test.
The fact that right after I asked if it was AI, a typical forgivable human error occurred. She forgot to paste the link. No worries there.
Except, that’s also the same error Assistant has on my phone(Pixel), where I ask it something and it just sits there not returning anything as if it believes it completed.
Then there’s the few forgivable human typos. Still fully able to comprehend what is being said so no harm no foul. The inaccurate tense, oh signs of typical foreign call center conversation. Except when she says "I wanted to help you with this", it doesn’t shout ‘foreign’ to me it shouts Google Assistant.
"I will consult your account… and let me try my best to…. seek approval to uplift the suspension" reads to me like a very very intelligent AI understanding and communicating.
but the final thing for me was "It can be, because it will be in a Zip file." after a five minute wait. At first I thought, okay. The person isn’t a tech and what I just said flew way over their head. but when I think about the question, it was a "Can this be done" question, through the form of "Is it possible."
The response goes, ‘yes. It can be. Because, definitive reason.’
But it goes on. The troubleshooting. "Did you?" ‘I hadn’t.’ "That is the reason it is not working."
Now that’s all of the literal red flags that set off for me. But for the whole conversation, I can see this perfectly logical thinking process involved. How it can see how one issue must be resolved before returning to a larger problem. Alright maybe that’s the minor psychosis I’m experiencing talking and maybe this poor soul had to deal with me.
..but here’s the final kicker.
Remove everything I just said and this still applies. During this conversation I felt instinctively that I was talking to a machine. That I was being tricked. That this was pretending to be a person. Not just the timing of the responses. Hidden cues. Not anything you can point a finger at. Just, that’s not a person.
So call me Orwellian(real word btw) here, but no one could possibly be anywhere close to this level of AI, that can literally think, literally find the answer to all your questions right? But what reason do I have to believe that Google’s early phase Duplex technology or even Google Assistant is the height of their prowess. Did they not learn from Glass not to rush things? Why would they show off the height of their power, their literal weapon at its full potential just to get some, what, investors? They wouldn’t. Assistant is already leagues ahead of Siri and Cortana. There’d be no competition if they realize the game was already won, that Google already made it to the Moon.
Think about it. Turing tests were passed decades ago. How do you evolve? You mass deploy. Where’s the best place, the only place to deploy your beyond state of the art tech? Your own platform.
from Artificial Intelligence https://ift.tt/2u7mLqU