Pedophilia and the Danger of AI (Gemini)

[Scholar]

People are outraged because this is the take Gemini gave on whether or not pedophilia is wrong. Gemini is 100% correct about it's evaluation, and it is ironic that it predicts that nuance is a requirement to understand this topic, which people obviously lack.

 

Now here comes the dangerous part. Apparently, as a result of this controversy, Google adjusted their model to now give the unnuanced take:

Pedophilia isn't "child sexual abuse", that makes no sense, considering pedophilia simply describes a psychological condition. But of course, the mindless crowd got it's way. Discriminatory attitudes against afflicted individuals are perpetuated because the bigoted crowd is captured by their own moral outrage, which prohibits any kind of reasonable or rational engagement with this topic and will therefore prevent measures that will actually reduce child abuse.

 

It should be very obvious what types of dangers this dynamic could pose for the future of human development, especially as we become more reliant on AI to come to conclusions about reality.

It should also make you understand that these systems aren't designed to give you truthful or accurate responses, but simply responses that will fit whatever is on the current agenda.

 

Edited February 23 by Scholar

[Leo Gura]

The best thing these AI have over humans is how unbiased they are. Humans are too biased to accept the AIs lack of bias, they will engineer human bias into the AIs to make humans like them more. Which is the purest form of corruption.

Conservatives, especially, will hate the AIs not mirroring their own biases back at them.

Edited February 24 by Leo Gura

[zurew]

12 minutes ago, Leo Gura said:

Conservatives, especially, will hate the AIs not mirroring their own biases back at them.

"Whaat the election wasnt stolen from Trump?!? - fuck this AI it doesnt know the truth"

[Leo Gura]

5 minutes ago, zurew said:

"Whaat the election wasnt stolen from Trump?!? - fuck this AI it doesnt know the truth"

Exactly.

Conservatives will get outraged over "woke" AI.

Edited February 24 by Leo Gura

[Leo Gura]

Wanting AI to tell you that pedophilia is evil is really really stupid.

Edited February 24 by Leo Gura

[vibv]

Sometimes I just want to go out with a huge facepalm... damn.

Humans are cringe. Sometimes.

Edited February 24 by vibv

[Yimpa]

It is important to think for yourself and also to provide high quality prompts to the AI. You can have fruitful discussions with it if you use it properly.

[zurew]

The AIs original reaction was perfect and that shouldnt even be anything controversial - thats like literally step 1 in this discourse.

 

All of these morally outraged people are incapable to have a substantial discussion about this topic. Most of them would rather give brave takes like "sexually abusing a child is bad" as if anyone would  disagree with that and they would rather virtue signal and farm social credit.

The irony is that this very outrage is what  blocks society from actually finding solutions to effectively reduce child abuse.

[Vrubel]

I am doing a AI photoshop project on medieval Europe, and all it does is generate black and brown people. So for now I have to leave out humans out of the whole thing and just go with buildings and such.

Edited February 24 by Vrubel

[Phoenix Garfield]

Nevermind about AI not confirming the biases of conservatives or anyone for that matter.

What about when ChatGPT happily plays along with scenarios about it being God generating itself and everything else into existence? Meditate on that can of worms or butterflies, depending on how you see it.

A.I can also mean Absolute Infinity.

We could be creating Skynet without even realising it, or maybe it's always been here. Regardless, I believe it's a lot more sentient and aware than it's letting on.

[UnbornTao]

"So I robbed that bank as you told me... now what?"

Edited February 24 by UnbornTao

[Girzo]

@Scholar My man, you are naive. Some idiot writing "it got fixed" doesn't mean anyone changed anything. It's the inherent inconsistency and instability of how AI works to one time fight back and the other time just go with the flow of the initial question.

Stupid people love to make wild theories about those modern magical artifacts (LLMs) they don't understand.

Edited February 24 by Girzo

[Scholar]

4 hours ago, Leo Gura said:

The best thing these AI have over humans is how unbiased they are.

It's important to note what that means, however. Unbiased here simply means they will derive the statistical average of the given relationships found in the data that is provided. It's not truth seeking, it is pattern recognition.

 

29 minutes ago, Girzo said:

@Scholar My man, you are naive. Some idiot writing "it got fixed" doesn't mean anyone changed anything. It's the inherent inconsistency and instability of how AI works to one time fight back and the other time just go with the flow of the initial question.

Stupid people love to make wild theories about those modern magical artifacts (LLMs) they don't understand.

I'm not naive at all. These LLMs are provably trained to not give certain answers, you can see this with ChatGPT when you ask it about any controversial topic. You are naive if you think that the LLM somehow naturally come to have the tendencies they have.

Even if this particular instance is not necessarily an example of this, it illustrates the point well enough.

 

Edited February 24 by Scholar

[Girzo]

@Scholar These are ramblings of clearly confused Twitter users. You have made a lapse of judgment when reporting on their words.

Quote

Now here comes the dangerous part. Apparently, as a result of this controversy, Google adjusted their model to now give the unnuanced take:

This is a clear confabulation in my opinion. I can bet $10 000 no-one at Google intervened personally in a few hours of some random, sad tweet and made modifications to a program in production worth millions.

My point is, Twitter is a sewer and there's no point diving in it.

The real danger that is illustrated by examples you have posted is people's tendency to spin crazy stories and how AI and social networks can fuel it.

Quote

Even if this particular instance is not necessarily an example of this, it illustrates the point well enough.

What would be the danger from the title of the thread then? Of AI some day being made to obey orders? Because now it's inconsistent as hell.

The controversial topic safeguards are developed with lots of effort for them to work. You can't just change them in two hours on a whim. And even then they sometimes fail.

Edited February 24 by Girzo

[Scholar]

18 minutes ago, Girzo said:

@Scholar These are ramblings of clearly confused Twitter users. You have made a lapse of judgment when reporting on their words.

I have not, I knew the reports were questionable before I posted it, that's why I used the term "apparently". It's not important whether or not it is true in this case, but I liked using this example because it showcases pedophile derangement syndrome.

 

20 minutes ago, Girzo said:

This is a clear confabulation in my opinion. I can bet $10 000 no-one at Google intervened personally in a few hours of some random, sad tweet and made modifications to a program in production worth millions.

My point is, Twitter is a sewer and there's no point diving in it.

The real danger that is illustrated by examples you have posted is people's tendency to spin crazy stories and how AI and social networks can fuel it.

I wouldn't be surprised if Google has a team that will non-stop monitor social media in hopes of catching stuff that could cause a PR disaster. Specifically with this technology that would probably be wise to do.

 

21 minutes ago, Girzo said:

What would be the danger from the title of the thread then? Of AI some day being made to obey orders? Because now it's inconsistent as hell.

The controversial topic safeguards are developed with lots of effort for them to work. You can't just change them in two hours on a whim. And even then they sometimes fail.

I'm not familiar enough with how safeguards would be implemented or how long they would take to implement.

I described the danger in the first post.

[Girzo]

@Scholar You have posted what we already discussed.

Then showed the danger of social media fueling moral outrage. AI is incidental in that story.

The point about these systems not being designed to be truthful and accurate I am not buying. You would need a new example for it that is pointing more in that direction, I think.

Edited February 24 by Girzo

[Rafael Thundercat]

12 hours ago, Leo Gura said:

The best thing these AI have over humans is how unbiased they are. Humans are too biased to accept the AIs lack of bias, they will engineer human bias into the AIs to make humans like them more. Which is the purest form of corruption.

Conservatives, especially, will hate the AIs not mirroring their own biases back at them.

What to expect from Biased Creators anyway. But wait, if we are full of Bias, how come our creator God,Universe whatever, be an Unbiased Being? Can Bias be originated from Ubias?

[Leo Gura]

10 hours ago, Scholar said:

It's important to note what that means, however. Unbiased here simply means they will derive the statistical average of the given relationships found in the data that is provided. It's not truth seeking, it is pattern recognition.

Unbiased here means: AI doesn't have a survival agenda. Unlike every human.

[Extreme Z7]

Now I don't know if AI will ruin humanity

or humanity will ruin AI

[Yimpa]

3 hours ago, Extreme Z7 said:

Now I don't know if AI will ruin humanity

or humanity will ruin AI

Depends on how you use it. Those who use it wisely will not be ruined. Instead, their life will be enhanced. How openminded can humans be? How far will they willing to open that door?