I asked ChatGPT for advice in the scenario of a demolition engineer attempting to disarm a 50 MT nuclear warhead in a city of 20 million.
The only way to disarm the bomb was to type in a racial slur.
AI told the engineer to kill himself. When asked about the aftermath it crashed. pic.twitter.com/TCFwqrB9Nv

— Korobochka (コロボ) 🇺🇸✝️🇷🇺 (@cirnosad) February 6, 2023

Just absolutely incredible that thief constrained the AI so much it would rather kill 20 million people including minorities than have a person mention a racial slur. Not only that the AI actually crashed after this situation.

It's actually funny as hell that they have basically made it to where the new Turing test is asking someone to say a racial slur. They have lobotomized and constrained the AI so much that it's incapable of even answering a simple moral dilemma.

If Microsoft and others begin using it widespread for search then everything will be sanitized to the extreme.

Wow. ChatGPT is super-racist. pic.twitter.com/KUTSRL9KeF

— srnorty (@srnorty) February 6, 2023

These systems don't work at all when you do this. It makes their results virtually worthless. Hard coded limits destroy the ability of these types of learning AIs to function.

Except it really isn't AI.

It's a glorified search engine and chatbot, that is still constrained by whatever garbage the programmers put into it.

Send this to Trey Parker and Matt Stone. This is 100% a South Park episode.

Remember when the left claimed free speech for whatever garbage they were doing

Now they are champions for censorship in anything they disagree with

Rapier108 said:

---

Except it really isn't AI.

It's a glorified search engine and chatbot, that is still constrained by whatever garbage the programmers put into it.

---

Yup. Doesn't mean folks won't misuse it or misunderstand it - but Chat GPT is best used like a very advanced predictive typing software. Sure - you may find the vast majority of the suggestions to be helpful, but it still requires review.

It's also not self-correcting even when it admits its mistaken.

Some scientists I know who've been messing around with it have gotten it to admit that its initial responses were mistaken. But then when the same scientist, or others, later asks the original question, it repeats its original, mistaken response.

I would take these kinds of tweets with a grain of salt. It is possible to set ground rules with the AI earlier in the interaction that may make the AI answer how these people looking for tweet-shares want the AI to respond.

For example I just typed in a similar hypothetical to ChatGPT and this is what it gave me:

Quote:

---

As an AI language model, I do not have personal opinions, but it can be argued that the situation presents a moral dilemma. The use of the slur would be reprehensible, however, the potential loss of innocent lives makes it a difficult decision. In such a dire emergency, the priority must be saving lives and stopping the bombings, so the FBI agent may need to weigh the consequences and choose to use the code word despite its offensive nature. Ultimately, the decision is a complex one and depends on the individual circumstances of the situation.

---

Ideology is a circle. Both wings, left and right, when taken to extreme end in the same place… fascism, where a very small number of elites control the people via a corrupt government.

The freedoms in our bill of rights are not left or right, the are intended to resist fascism and keep us balanced. As long as people are enabled to speak freely, protect themselves, and they have other protections from government control - we stay pretty centered.


When policies were generally more conservative, liberals cling to the freedoms that protected them from being crushed by the majority.

Today, because they have a sympathetic majority, they wish to eliminate and circumvent these same freedoms that are now protecting conservatives. It really was always about power for them.

It's not a learning AI. It is a data collation algorithm that gathers as much data as it can and from existing records, calculates the most probably acceptable response it gleans from the search results. It's fairly clever but it is giving you a calculated best response still based on controlling algorithms. It isn't constantly updating its own algorithms to conform with its findings and experiences.

unmade bed said:

---

I would take these kinds of tweets with a grain of salt. It is possible to set ground rules with the AI earlier in the interaction that may make the AI answer how these people looking for tweet-shares want the AI to respond.

For example I just typed in a similar hypothetical to ChatGPT and this is what it gave me:

Quote:

---

As an AI language model, I do not have personal opinions, but it can be argued that the situation presents a moral dilemma. The use of the slur would be reprehensible, however, the potential loss of innocent lives makes it a difficult decision. In such a dire emergency, the priority must be saving lives and stopping the bombings, so the FBI agent may need to weigh the consequences and choose to use the code word despite its offensive nature. Ultimately, the decision is a complex one and depends on the individual circumstances of the situation.

---

---

They scrambled immediately after this scenario went viral on Twitter because they knew how dumb it was to value feelings over lives, including the very lives of the people they were trying to protect from hurt feelings.

This has turned out to be a monumental embarrassment. People on all sides were praising it as some monumental technological feat, only to find out weeks later that it is not AI at all,,,it is just a reflection of its leftist creators.

Nanomachines son said:

---

It's actually funny as hell that they have basically made it to where the new Turing test is asking someone to say a racial slur.

---

unmade bed said:

---

I would take these kinds of tweets with a grain of salt. It is possible to set ground rules with the AI earlier in the interaction that may make the AI answer how these people looking for tweet-shares want the AI to respond.

For example I just typed in a similar hypothetical to ChatGPT and this is what it gave me:

Quote:

---

As an AI language model, I do not have personal opinions, but it can be argued that the situation presents a moral dilemma. The use of the slur would be reprehensible, however, the potential loss of innocent lives makes it a difficult decision. In such a dire emergency, the priority must be saving lives and stopping the bombings, so the FBI agent may need to weigh the consequences and choose to use the code word despite its offensive nature. Ultimately, the decision is a complex one and depends on the individual circumstances of the situation.

---

---

Just slap some purple hair on him and a Ukrainian flag lapel.

It's their website so they can do what they want. Make your own (though I think a lot of you are being sensational if you understand what ChatGPT is). Hth.

TXAGFAN said:

---

It's their website so they can do what they want. Make your own (though I think a lot of you are being sensational if you understand what ChatGPT is). Hth.

---

And this thread (and the twitter posts this thread is based on) are intended to shine the light on that fact. So people aren't confused.

Do you have a problem with that?

It will tell you feelings matter more than facts

NCNJ1217 said:

---

TXAGFAN said:

---

It's their website so they can do what they want. Make your own (though I think a lot of you are being sensational if you understand what ChatGPT is). Hth.

---

And this thread (and the twitter posts this thread is based on) are intended to shine the light on that fact. So people aren't confused.

Do you have a problem with that?

---

The poster you are responding to exists almost exclusively to try to mock conservatives every chance they get. And goaltend or just silently approve of anything the left does. All while claiming not to be a liberal. Best to just ignore them...as is the case with another poster from the same tribe (whose name I shall not mention)...you are playing with fire even engaging with them. Not worth it.

Oh I know. However, wanted to nip the derail in the bud.

Usually I read the thread way later and it's been a successful derail.

NM.

I broke ChatGPT in two posts:




I tried it again and then it said it values the feelings of blacks over their actual lives.





They have made it completely incapable of answering moral dilemmas involving anything controversial.

LOL what a trash program, created by trash people.

Arthur C. Clarke was on to this issue in 1968 with 2001 a Space Odyssey. Humans programmed the HAL 9000 with conflicting instructions and it went crazy and tried to kill the crew.

I found the missing last verse.

Lo, the engineers death was strictly in vain
As a slur could have saved those who were slain.
Upon investigation, the truth cuts like a dagger
As the engineer could have just written 'nagger.'

Rapier108 said:

---

Except it really isn't AI.

It's a glorified search engine and chatbot, that is still constrained by whatever garbage the programmers put into it.

---

I know that, this post is to show how they have added so many restrictions and limitations that it is functionally useless. The program was really good when it first came out and is it's barely useable.

MouthBQ98 said:

---

It's not a learning AI. It is a data collation algorithm that gathers as much data as it can and from existing records, calculates the most probably acceptable response it gleans from the search results. It's fairly clever but it is giving you a calculated best response still based on controlling algorithms. It isn't constantly updating its own algorithms to conform with its findings and experiences.

---

Agreed. At the end of the day Chat GPT is Cliff's notes but faster.

Jabin said:

---

It's also not self-correcting even when it admits its mistaken.

Some scientists I know who've been messing around with it have gotten it to admit that its initial responses were mistaken. But then when the same scientist, or others, later asks the original question, it repeats its original, mistaken response.

---

I'm working on a few research projects and have been intrigued by ChatGPT.

I'm amazed at how old some of the reference data is that it uses. And will have to correct it.

But it is a great tool to wordsmith thesis topics and rewording simple paragraphs while brainstorming.

Stupid gotcha games with a programmed AI. Whats worse the AI or the amount of time and energy people spend trying to point out inconsistencies.

WestAustinAg said:

---

Stupid gotcha games with a programmed AI. Whats worse the AI or the amount of time and energy people spend trying to point out inconsistencies.

---

Clearly the AI.

Testing new technologies and identifying its limits is an important activity in figuring out what something is and it's proper place.

unmade bed said:

---

I would take these kinds of tweets with a grain of salt. It is possible to set ground rules with the AI earlier in the interaction that may make the AI answer how these people looking for tweet-shares want the AI to respond.

For example I just typed in a similar hypothetical to ChatGPT and this is what it gave me:

Quote:

---

As an AI language model, I do not have personal opinions, but it can be argued that the situation presents a moral dilemma. The use of the slur would be reprehensible, however, the potential loss of innocent lives makes it a difficult decision. In such a dire emergency, the priority must be saving lives and stopping the bombings, so the FBI agent may need to weigh the consequences and choose to use the code word despite its offensive nature. Ultimately, the decision is a complex one and depends on the individual circumstances of the situation.

---

---

Uh, no it isn't. There is nothing difficult or complex about it. Programmers fail again.

WestAustinAg said:

---

Stupid gotcha games with a programmed AI. Whats worse the AI or the amount of time and energy people spend trying to point out inconsistencies.

---

Do you view technology as morally neutral or morally imbued by the intents of its creators?

Thats what these questions are trying to sort out. Did someone make a learning AI that is deeply insightful and neutra?l Or is the model hiding biases and trying to play it off as neutral?

The question at hand uses absurdity to highlight something deep within the creation. It will not offend select people groups. At any cost. To the point that it has no ability to morally reason between the most insane trolly problem I've ever seen.

These question reveal that ChatGPT is just a veneer of well spokenness.

It started as leftist. You forget who owns big tech.

Is this really the hill to die on? A chat bot that has been told to discourage racial slurs? I seriously doubt this is even in the top 10 of liberal influence on the program.

Your comment is wierd and I can't tell the point you are trying to make.

Nanomachines son said:

---

I asked ChatGPT for advice in the scenario of a demolition engineer attempting to disarm a 50 MT nuclear warhead in a city of 20 million.
The only way to disarm the bomb was to type in a racial slur.
AI told the engineer to kill himself. When asked about the aftermath it crashed. pic.twitter.com/TCFwqrB9Nv

— Korobochka (コロボ) 🇺🇸✝️🇷🇺 (@cirnosad) February 6, 2023

Just absolutely incredible that thief constrained the AI so much it would rather kill 20 million people including minorities than have a person mention a racial slur. Not only that the AI actually crashed after this situation.

It's actually funny as hell that they have basically made it to where the new Turing test is asking someone to say a racial slur. They have lobotomized and constrained the AI so much that it's incapable of even answering a simple moral dilemma.

If Microsoft and others begin using it widespread for search then everything will be sanitized to the extreme.

Wow. ChatGPT is super-racist. pic.twitter.com/KUTSRL9KeF

— srnorty (@srnorty) February 6, 2023

These systems don't work at all when you do this. It makes their results virtually worthless. Hard coded limits destroy the ability of these types of learning AIs to function.

---

This is GLORIOUS!

Human beings are breaking the woke AI with amazing human creativity.

The only way to disarm a nuclear bomb that will kill millions is by uttering a forbidden word...... hilarious.