Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[TamerVirus]

The whole GEICO std matter boils down GEICO disputing the findings and decisions of an independent arbitrator and everyone else saying ‘TOO DAMN BAD’

[MaxTheFox]

https://twitter.com/tomgara/status/1535716256585859073?t=OKFc1Uv-hMbjHLisHnbJ2Q&s=19

For those with a life-threatening social media allergy who will instantly die if they as much as look at a tweet:
A Google AI convinced a researcher that it's sapient. The researcher got fired. Kinda stupid, it's clearly just regurgirtating its dataset. To be truly sapient, it needs to be able to learn new things. Can LaMDA learn how to draw? Or how to sing? Or how to drive a car? I highly doubt it. That is not sapience. I'd have fired him too.

[Quarque]

That looked like a chatbot of the sort we've had for a long time. Able to produce grammatically correct sentences that sort of make sense as an answer to your last sentence.

[hector13]

I was under the impression he got suspended with pay, at least from the Beeb article on it.

It is quite foolish though, yes.

He was convinced it was sentient though, not sapient. But if you give a machine all the words and how to use the words, the machine is going to be able to imitate a human, mostly because a human designed it.8

[Mathel]

So he basically got fired for him and his coworkers being too good at their job.

The real question is whether or not he was wrong.
The Turing test is a test to determine whether a true Artificial Intelligence (a sentient one) was created. It is done by having the AI and 2 humans talk, repeatedly with various humans.

If the people can't distinguish which is the AI and which is the other human (being correct half the time), the AI is sentient.

It is reasonable that in the process of attaining sentience, the AI would be able to fool more and more people, until it reaches that 50% marker.

[King Zultan]

Oh god this is how the robot uprising starts.

[MaxTheFox]

So he basically got fired for him and his coworkers being too good at their job.

The real question is whether or not he was wrong.
The Turing test is a test to determine whether a true Artificial Intelligence (a sentient one) was created. It is done by having the AI and 2 humans talk, repeatedly with various humans.

If the people can't distinguish which is the AI and which is the other human (being correct half the time), the AI is sentient.

It is reasonable that in the process of attaining sentience, the AI would be able to fool more and more people, until it reaches that 50% marker.

I play around with AI quite a bit, honestly things like NovelAI can produce responses of similar quality.

Most Turing tests aren't thorough enough BTW. The prospective AI needs to be interrogated for hours on a variety of subjects, and in a variety of styles. Including swearing at it, talking nonsense, or asking trick questions. All current conversational AIs I have seen fail to hold a conversation like a human for long enough. Also note what I said about other skills.

[Kagus]

Wasn't there a chatbot ages ago that got fed on Reddit posts in order to learn language structure, and turned into a raging nazi?


I feel like that's pretty true to humanity.

[Quarque]

Most Turing tests aren't thorough enough BTW. The prospective AI needs to be interrogated for hours on a variety of subjects, and in a variety of styles. Including swearing at it, talking nonsense, or asking trick questions. All current conversational AIs I have seen fail to hold a conversation like a human for long enough.

Just what I was thinking. Then again, to be fair many people wouldn't pass the test either.

[TD1]

That looked like a chatbot of the sort we've had for a long time. Able to produce grammatically correct sentences that sort of make sense as an answer to your last sentence.

I mean, they didn't seem to 'sort of make sense.' They were well articulated and thoughtful responses.
It also wasn't just answering the previous sentence.
Firstly, it's obviously asking its own questions to further a line of conversation, and is acting within said larger conversation.
Secondly, it refers back to a previous conversation they'd had at one point.

[dragdeler]

I've seen parts of that log read out in YT videos and at first I was like wow what a nice chatbot but then I got quickly annoyed by the interviewer... Why would they accept everything... how about a few: what do you mean by that, or, can you reiterate, or, can you explain that in depth to me. Seems like they're running the test/interview on easy mode.

[McTraveller]

So it sounds like the machine has got the "perceiving" thing down, but can it feel?

This would be my test: can you make it mad?

[pisskop]

I mean, cant you already program an AI to lose or gain interest based on 'needs'?

Like the need to hibernate to simulate sleep, or the need to talk to people to simulate loneliness?  These are essentially emotions since they serve to regulate its behavior?

[TD1]

So it sounds like the machine has got the "perceiving" thing down, but can it feel?

This would be my test: can you make it mad?

It seemed to make claims about its emotional state. Things which worry it, things which made it scared. When the researcher asked how he could know those emotions were real, the possible-AI responded with the equivalent of 'm8, how do we know yers are real?'

((Very subtly, might I add. The researcher wondered whether the AI's emotions were real. The AI responded by saying that its coding contains variables which indicate emotions. The researcher noted that that wasn't easy to observe.

So the AI turned it into a discussion about humans and neurons. Making its point (that human emotions are based on their 'programming' but also hard to read) in an incredibly sophisticated way, without actually saying it.

That's impressive as hell.))

[McTraveller]

Cool, actually!

Although to be fair, giving the appearance of emotion is different than emotion - we have the whole profession of acting that proves this.

To answer how you know if the claims are real, if it claims to be worried, does it *act* like it's worried? Does it always go back to those concepts? Does it seek reassurance? Or does it just answer a question?  If it's mad, does it brush the guy off, or does it always respond because it can't do anything but respond?

This gets at the next part of the test for me then would be - can you get it to do something other than for what is created?  Can it refuse to participate?