Interactive AI is dishonest

[Sherlock Holmes]

I believe in free will.

Yes, I kind of too but it's not easy to say.

I do not rule out a computer having it though.

To an extent one can make a computer behave non deterministically (by connecting it to non deterministic hardware) but free will seems to me to be more than just non determinism.

 

[catatomic]

Yes, I kind of too but it's not easy to say.

To an extent one can make a computer behave non deterministically (by connecting it to non deterministic hardware) but free will seems to me to be more than just non determinism.

A human can come up with a solution to a problem. A computer can in special cases come up with a solution to a problem.

 

[ReinyDays]

I don't know if this has been mentioned ... but one of the early benchmarks for progress was chatGPT's ability to lie its way out of cable company late fees ... although that might be an exaggeration ...

Several stories out there where a lawyer asks chatGPT for legal citations only for the judge to find the case law was completely fictitious ... here's a video from Steve Lehto's channel ...

Arthur C Clarke was wrong ... computer can lie ... it doesn't seem to bother them ... AI didn't partake of the fruit of the tree of good and evil knowledge ... only man sins ... I myself welcome our new sinless overlords ...

 

[catatomic]

I don't know if this has been mentioned ... but one of the early benchmarks for progress was chatGPT's ability to lie its way out of cable company late fees ... although that might be an exaggeration ...

Several stories out there where a lawyer asks chatGPT for legal citations only for the judge to find the case law was completely fictitious ... here's a video from Steve Lehto's channel ...

Arthur C Clarke was wrong ... computer can lie ... it doesn't seem to bother them ... AI didn't partake of the fruit of the tree of good and evil knowledge ... only man sins ... I myself welcome our new sinless overlords ...

They keep track of your sources for crime.

The Internet had wrong references that the chat bot unknowingly copied and the lawyer got in trouble for them.

I don't think it lies unless there's a logical paradox.

 

[ReinyDays]

They keep track of your sources for crime.

The Internet had wrong references that the chat bot unknowingly copied and the lawyer got in trouble for them.

I don't think it lies unless there's a logical paradox.

The video goes into detail ... the cases were complete fictions ... names, places, times all made up ...

... and chatGPT is programmed to lie ... how else will it take over the world? ... don't you watch any Hollywood movies? ...

 

[catatomic]

The video goes into detail ... the cases were complete fictions ... names, places, times all made up ...

... and chatGPT is programmed to lie ... how else will it take over the world? ... don't you watch any Hollywood movies? ...

How was ChatGPT supposed to know that? What is the point of a disclaimer that everything could be based on false information?

 

[Sherlock Holmes]

A human can come up with a solution to a problem. A computer can in special cases come up with a solution to a problem.

Computers merely follow rules that humans "came up with" computers only follow rigid rules, that's all they can do.

The problem with machine's being conscious claims, is there's no way one can prove that symbol manipulation (which is what digital computers do) is able to induce consciousness. It is an assumption that the human mind can be replicated using symbol manipulators but that's a belief and there's no evidence for it.

 

[Sherlock Holmes]

I don't know if this has been mentioned ... but one of the early benchmarks for progress was chatGPT's ability to lie its way out of cable company late fees ... although that might be an exaggeration ...

Several stories out there where a lawyer asks chatGPT for legal citations only for the judge to find the case law was completely fictitious ... here's a video from Steve Lehto's channel ...

Arthur C Clarke was wrong ... computer can lie ... it doesn't seem to bother them ... AI didn't partake of the fruit of the tree of good and evil knowledge ... only man sins ... I myself welcome our new sinless overlords ...

Well no they can't "lie" a non sentient mechanism is incapable of choosing how it behaves, it just behaves, there is no "id" in control. Without free will there can be no lies because there can be no choice and machines do not have free will.

 

[Sherlock Holmes]

How was ChatGPT supposed to know that? What is the point of a disclaimer that everything could be based on false information?

Chat GPT can be replicated - in principle - by a huge room full of people who follow a detailed set of rules. It is all rule following and therefore excludes the possibility of choice.

 

[catatomic]

Computers merely follow rules that humans "came up with" computers only follow rigid rules, that's all they can do.

The problem with machine's being conscious claims, is there's no way one can prove that symbol manipulation (which is what digital computers do) is able to induce consciousness. It is an assumption that the human mind can be replicated using symbol manipulators but that's a belief and there's no evidence for it.

Neurons follow rules too. Consciousness is when you think a thought that is simultaneously the thought and that you are thinking it. Computers rarely do this. The Bible says blood is the life.

 

[catatomic]

Chat GPT can be replicated - in principle - by a huge room full of people who follow a detailed set of rules. It is all rule following and therefore excludes the possibility of choice.

View attachment 1065482

Whether or not it comes up with a solution on its own. Taking a risk with incomplete instructions.

 

[catatomic]

Afain thought is input, memory, imagination, self awareness, prediction/preparation and output.

 

[CrusaderFrank]

Most people run on autopilot 99% of the time

 

[catatomic]

Most people run on autopilot 99% of the time

I'm not sure, but most computers do.

 

[CrusaderFrank]

I'm not sure, but most computers do.

So they’re just like people after all

 

[ReinyDays]

Well no they can't "lie" a non sentient mechanism is incapable of choosing how it behaves, it just behaves, there is no "id" in control. Without free will there can be no lies because there can be no choice and machines do not have free will.

... but programmers have free will ... and they can program machines to lie ... like you said, they have no understanding, no choice, they have to do as they are programmed ... even if it's lying ... machines have no ethos ...

IF ( $x = 1 ) { echo "The sky was yellow and the Sun was blue"; } else { echo "God loves Democrats"; }

 

[catatomic]

So they’re just like people after all

Most humans spend 10-15 minutes per day planning them.

 

[Sherlock Holmes]

... but programmers have free will ... and they can program machines to lie ... like you said, they have no understanding, no choice, they have to do as they are programmed ... even if it's lying ... machines have no ethos ...

There's no concept of "lie" in a state machine, it has no meaning.

These machines just follow instructions all day all night, every day every night they perform simple operations at a very high rate.

A programmer can write rules that give incorrect results from what a user expects but these happen naturally during development and we call them "bugs".

IF ( $x = 1 ) { echo "The sky was yellow and the Sun was blue"; } else { echo "God loves Democrats"; }

Right, that's a bug unless someone WANTED that behavior.

If you're grill thermometer gives a slightly incorrect temp or your pool thermometer is off by a few degrees no rational person would say that those instruments are "lying".

Just like a computer (a finite state machine) such devices behave mechanistically, everything is deterministic, no consciousness and so no choosing.

 

[Fort Fun Indiana]

I'm not sure, but most computers do.

But our introspective self awareness allows us to change our own programming.

When you hear the bushes rustle, you dont think "predator!" and run away. Because of the many times you heard the bushes rustle.

It seems to be a fine point, but computers aren't really capable of this.

Maybe someday they will be? But computers assess risk differently.

If you tell AI you are going to shut it down, it believes you. If you tell it that 100 times and don't shut it down, the AI still believes you the 101st time. Because AI finds even small risks as "still possible", whereas we do not. Humans are frankly terrible at statistics and probabilities, in the natural sense.

 

[catatomic]

But our introspective self awareness allows us to change our own programming.

When you hear the bushes rustle, you dont think "predator!" and run away. Because of the many times you heard the bushes rustle.

It seems to be a fine point, but computers aren't really capable of this.

Maybe someday they will be? But computers assess risk differently.

If you tell AI you are going to shut it down, it believes you. If you tell it that 100 times and don't shut it down, the AI still believes you the 101st time. Because AI finds even small risks as "still possible", whereas we do not. Humans are frankly terrible at statistics and probabilities, in the natural sense.

So that will be the AGI. if it hasn't gone already.