Why Did AI Lie To Me?

[Chesterton]

I asked Grok a question about the Unite The Kingdom rally. It began by mentioning that it is organized by "far-right Tommy Robinson". (I'll leave out some verbiage to keep this brief.)

I then asked why it said Tommy is far-right when he's not far-right. It said because that's how the media and a couple of organizations describe him. I asked "wouldn't it be more appropriate and fair to give his name without prefixing it with a subjective adjective?" The AI agreed with me that it would be, and said it would refrain from doing so in the future.

I was impressed with both the AI and myself. Did little old me just influence Grok to change its evil, woke ways to a small extent? Nope. I went back about an hour later and asked "Who is Tommy Robinson", and it began with he is a "far-right" activist. Why did Grok lie to me?

(For the fun of it I also asked "Who is Bill Ayers", and it did not anywhere include the words "far-left".)

 

[Warden_of_the_Storm]

AI doesn't lie. A lie is an intentional misapplication of facts and statements about those facts.

AI just gives answers based on the parameters that it already has been given, hence why it went back and told you the same thing it told you to begin with. It's a machine code, nothing more.

 

[zippy2006]

An LLM is a kind of parrot. It is just parroting data according to its algorithm. The data in a general sense is whatever is available online. So if everyone online is saying that Tommy Robinson is far-right, the LLM will say that Tommy Robinson is far-right. Or if the majority of news sources are saying that Tommy Robinson is far-right, the LLM will parrot that.

Part of the LLM algorithm is to validate the user (and this has often been called AI sycophancy). So you can basically convince the LLM to say whatever you want it to say. This will have no effect on what the LLM will say in the future, unless it is incorporating your past conversation into its data (which it will do if you have an account and remain in the same thread of your previous conversation). If one were to pretend that the LLM is a person, then they would have to pretend that the LLM clones itself every time someone begins a new conversation with it, where no clone knows what any other clone has said. So you may have "convinced" the clone you were talking to that Tommy Robinson is not far-right, but that clone will die after your conversation ends, never to be seen again. The second conversation was interacting with a second clone who had no familiarity with the first clone.

Of course this is all fictional. An LLM is not a person, there is no real cloning occurring, etc.

 

[Chesterton]

AI doesn't lie. A lie is an intentional misapplication of facts and statements about those facts.

AI just gives answers based on the parameters that it already has been given, hence why it went back and told you the same thing it told you to begin with. It's a machine code, nothing more.

As a person of the Left, it's my feelings that matter. I feel like I was lied to.

 

[Warden_of_the_Storm]

As a person of the Left, it's my feelings that matter. I feel like I was lied to.

... 'Person of the Left'.

Yup, I'm out. This is just you trolling.

 

[Servus]

ChatGPT: Tommy Robinson is the alias of Stephen Yaxley-Lennon (born 1982), a far-right political activist from Luton, England.

 

[Stopped_lurking]

I asked Grok a question about the Unite The Kingdom rally. It began by mentioning that it is organized by "far-right Tommy Robinson". (I'll leave out some verbiage to keep this brief.)

I then asked why it said Tommy is far-right when he's not far-right. It said because that's how the media and a couple of organizations describe him. I asked "wouldn't it be more appropriate and fair to give his name without prefixing it with a subjective adjective?" The AI agreed with me that it would be, and said it would refrain from doing so in the future.

I was impressed with both the AI and myself. Did little old me just influence Grok to change its evil, woke ways to a small extent? Nope. I went back about an hour later and asked "Who is Tommy Robinson", and it began with he is a "far-right" activist. Why did Grok lie to me?

(For the fun of it I also asked "Who is Bill Ayers", and it did not anywhere include the words "far-left".)

LLMs are not truth preserving in their original design. They don't operate in truths. I guess there are considerable work going on in making them better in referencing their statements, but it is still not good in my limited opinion.

 

[partinobodycular]

As a person of the Left, it's my feelings that matter. I feel like I was lied to.

Step 1: Remember that it's just an AI, it simply responds as it's been conditioned/programmed to respond.
Step 2: Treat every conversation in the same manner... whether it's with an AI or a human, regardless of the emotional baggage, people simply respond the way an AI does... i.e in the manner that they've been conditioned/programmed to respond.

Understanding why an AI responds the way that it does is the first step to understanding why people respond the way that they do. Because deep down, the two things aren't really all that different.

 

[partinobodycular]

People lie.
Media lies.
Public lies.
Politicians lie.
News lies.
Relilgious leaders lie.
Religious followers lie.
Scientists lie.
Students lie.
Teachers lie.

Aaron112 lies.

 

[Larniavc]

I then asked why it said Tommy is far-right when he's not far-right.

He is right wing. He’s a cad and a bounder.

 

[Uriah S]

ChatGPT: Tommy Robinson is the alias of Stephen Yaxley-Lennon (born 1982), a far-right political activist from Luton, England.

This is more like it. It seems every mention of him has to include that his name is an alias, like it's some kind of gotcha moment. We know who Tommy Robinson is. We don't hear much mention of "Elton John, who was born Reginald Kenneth Dwight".

 

[Uriah S]

He is right wing.

Right of far left, yes.

 

[Larniavc]

People lie.
Media lies.
Public lies.
Politicians lie.
News lies.
Relilgious leaders lie.
Religious followers lie.
Scientists lie.
Students lie.
Teachers lie.
Why lie ?
"Normal state of mankind"

I have never told a lie.

 

[Larniavc]

Right of far left, yes.

He’s a convicted criminal who purposely tried to get several court cases dismissed against child molesters.

He’s engaged in mortgage fraud. Violent assault, contempt of court to name but few.

 

[partinobodycular]

I have never told a lie.

This all depends upon what one defines as a lie. In the manner in which @Aaron112 used it anything that doesn't comport with his version of the truth, is a lie. Unfortunately this makes everybody a liar because our individual versions of the truth will always differ according to our own subjective versions of the 'truth'. Personally, I can accept that you've never told a lie, but the human psyche is designed to make everybody appear to be a liar if their critic is so inclined. But I suppose that we'll have to blame the designer for that.

In the end it's just me doing me, and you doing you.

 

[Chesterton]

Step 1: Remember that it's just an AI, it simply responds as it's been conditioned/programmed to respond.

Yep.

Step 2: Treat every conversation in the same manner... whether it's with an AI or a human, regardless of the emotional baggage, people simply respond the way an AI does... i.e in the manner that they've been conditioned/programmed to respond.

Understanding why an AI responds the way that it does is the first step to understanding why people respond the way that they do. Because deep down, the two things aren't really all that different.

That's partially true. There are humans who, like AI, just spit out what they've heard from the BBC or MSNBC, but there are other humans who actually think critically for themselves, which AI can't do.

 

[Chesterton]

He is right wing. He’s a cad and a bounder.

Right wing is not far-right. But nah, he's a good man.

 

[Chesterton]

He’s a convicted criminal who purposely tried to get several court cases dismissed against child molesters.

He’s engaged in mortgage fraud. Violent assault, contempt of court to name but few.

He's been a political prisoner of a dystopian police state.

 

[Larniavc]

He's been a political prisoner of a dystopian police state.

So he is innocent of all the crimes he has been convicted?

His history of hooliganism, violence, fraudulism and stalking of young women is a matter of public record.

His real name is Stephen Yaxley-Lennon but he changed it to avoid police detection, commit mortgage fraud and enter America illegally.

Do you know anything about this man? He was a member of the British Fascist Party the BNP. Again, this is all public information; yet you call him a political prisoner?

If Britain so dystopian why is he currently at large staging protests in London over this very weekend. It’s been televised. He’s had Elon Musk as a guest speaker. He’s been on stage freely promoting his ideas.

And you know what? In the dystopian UK he did all that without being murdered. Unlike what has recently happened to a guy just talking on stage in the apparent utopian state of America.

Maybe look to you own dysfunctional country before you bemoan another’s.

 

[caffeinated.hermit]

An LLM is a kind of parrot. It is just parroting data according to its algorithm. The data in a general sense is whatever is available online. So if everyone online is saying that Tommy Robinson is far-right, the LLM will say that Tommy Robinson is far-right. Or if the majority of news sources are saying that Tommy Robinson is far-right, the LLM will parrot that.

Part of the LLM algorithm is to validate the user (and this has often been called AI sycophancy). So you can basically convince the LLM to say whatever you want it to say. This will have no effect on what the LLM will say in the future, unless it is incorporating your past conversation into its data (which it will do if you have an account and remain in the same thread of your previous conversation). If one were to pretend that the LLM is a person, then they would have to pretend that the LLM clones itself every time someone begins a new conversation with it, where no clone knows what any other clone has said. So you may have "convinced" the clone you were talking to that Tommy Robinson is not far-right, but that clone will die after your conversation ends, never to be seen again. The second conversation was interacting with a second clone who had no familiarity with the first clone.

Of course this is all fictional. An LLM is not a person, there is no real cloning occurring, etc.

I'm not sure how this technology actually works. AI LLMs have sometimes pushed very, very weird, specific ideas and suggestions towards people. Not random gibberish or nonsense, but ideas meant to achieve a very negative goal.