The overconfident AI problem

[_TrXtR_]

The Chinese artificial intelligence app growing in popularity in the U.S. is unable to provide any information on the 1989 Tiananmen Square protests or criticize Chinese President Xi Jinping.

The inability to provide information on the 1989 Tiananmen Square protests could indicate that DeepSeek's ability to work as a search engine and information provider is compromised by its Chinese origins.

DeepSeek AI refuses to criticize Xi Jinping: "Talk about something else"

Chinese-owned DeepSeek AI was also unable to provide any information on Tiananmen Square when asked by Newsweek.

www.newsweek.com

Yeah AI is censored depending on the country/culture.
This is normal. I curious maybe Chinese version of Wikipedia might also not contain the info.

It's an LLM and yes it spews based on what it's trained. So you going to get what you want depending on how you train it. But being upset about how others train their AI makes it seem like you are rather against freedom of speech and anything thst is allowed is just what you approve of.

 

[_TrXtR_]

You can identify as a cat and meow all day. I'm not going to pretend you're a cat through.

Also lmao pretending this nonsense is correct, haha.

Don't let Ai control you man. Though I haven't seen an AI that promoted you or requirs to you to identify as a cat.
I think you're more confused than an lgbtqrx++ activist

 

[Swa]

Yeah AI is censored depending on the country/culture.
This is normal. I curious maybe Chinese version of Wikipedia might also not contain the info.

It's an LLM and yes it spews based on what it's trained. So you going to get what you want depending on how you train it. But being upset about how others train their AI makes it seem like you are rather against freedom of speech and anything thst is allowed is just what you approve of.

But that defeats the purpose. If it needs to be trained by people it's not AI.

 

[Cosmik Debris]

Yeah AI is censored depending on the country/culture.
This is normal.

You see the censorship and skewing of historical fact as normal?

But being upset about how others train their AI makes it seem like you are rather against freedom of speech and anything thst is allowed is just what you approve of.

So, according to your ethics, censorship (Omission of facts) counts as freedom of speech?

 

[_TrXtR_]

You see the censorship and skewing of historical fact as normal?



So, according to your ethics, censorship (Omission of facts) counts as freedom of speech?

There are 2 arguments that appeared here.
1] the bots are woke because it mentions gender identity
2) information being excluded because that government obviously doesn't like the information

2 conflicting arguments the one is against censorship the other isn't.
Freedom of choice/speech would include the censorship of speech.
If i want to write a book and exclude certain information and include other information isn't that exactly freedom of speech?
Don't use an AI you don't like. What so hard about that? You still have freedom of choice or are you being forced?

 

[Cosmik Debris]

Freedom of choice/speech would include the censorship of speech.

What? Any limitation of the freedom of speech renders such freedom null and void. Who decides on the limitation? The effects in the UK are already apparent. Criticism of Muslims or Protests on Facebook result in jail sentences.

If i want to write a book and exclude certain information and include other information isn't that exactly freedom of speech?

It's an unethical abuse of freedom of speech.

 

[_TrXtR_]

What? Any limitation of the freedom of speech renders such freedom null and void. Who decides on the limitation? The effects in the UK are already apparent. Criticism of Muslims or Protests on Facebook result in jail sentences.



It's an unethical abuse of freedom of speech.

Except when it includes things you don't want to see

 

[Cosmik Debris]

Except when it includes things you don't want to see

Such as?

 

[Swa]

There are 2 arguments that appeared here.
1] the bots are woke because it mentions gender identity
2) information being excluded because that government obviously doesn't like the information

2 conflicting arguments the one is against censorship the other isn't.
Freedom of choice/speech would include the censorship of speech.
If i want to write a book and exclude certain information and include other information isn't that exactly freedom of speech?
Don't use an AI you don't like. What so hard about that? You still have freedom of choice or are you being forced?

Let's just unpack this. Gender fluidity is a fringe concept. Why should an AI mention that then?

Then AI is supposed to learn by itself. Censorship becomes a null and void concept then if information is excluded because it's not important.

 

[Ninja'd]

They cannot even answer what is a man

A miserable pile of secrets. But enough talk, have at you!

 

[_TrXtR_]

What? Any limitation of the freedom of speech renders such freedom null and void. Who decides on the limitation? The effects in the UK are already apparent. Criticism of Muslims or Protests on Facebook result in jail sentences.



It's an unethical abuse of freedom of speech.

Who are you to decide what I put in my book? It's my freedom to put what I want in it. And you are welcome to read someone else's book.

 

[_TrXtR_]

Let's just unpack this. Gender fluidity is a fringe concept. Why should an AI mention that then?

Then AI is supposed to learn by itself. Censorship becomes a null and void concept then if information is excluded because it's not important.

You can't grt rid of censorship.
There are things that are illegal in some countries and in some not. This is the norm of the world.
If something is fringe it's still information. Just because you don't like a topic doesn't mean it would be useful to know some details.

All AI are not the same. Each one has to abide to laws and possibly culture.
In sure someone can create an AI that doesn't know anything about woke. No one is stopping anyone from doing that.

But then ultimately that means censoring anything that you don't agree with and covering topics that you do agree with.

It's like saying AI shouldn't know anything about flat earth theory. And then throw it into an argument with a flat farther and walla you have a useless AI.

Every country has their rules/laws. So yes that affects training and thst show it is supposed to be. So if you gonna use a Chinese AI you gonna be stuck to their laws/culture
You use an American one and you gonna learn about women that have penis.
If you use a Muslim one you gonna learn about marrying very young and not eating pork.

Expecting it to work any other way is silly.

It would be useless creating an AI that contradicts laws and then gets squashed.

 

[Cosmik Debris]

Who are you to decide what I put in my book? It's my freedom to put what I want in it. And you are welcome to read someone else's book.

Sure, you can put whatever you like in your book.

But then I have the right to criticise your book so heavily for not being factual that you will lose all credibility and never sell a book again except to the fringe market that still believe you.

Do you see how important ethics and honesty are to credibility now?

 

[Cosmik Debris]

All AI are not the same. Each one has to abide to laws and possibly culture.

Then it's not intelligent. Intelligence transcends laws and cultures by applying critical thinking. I seriously doubt AI could come up on it's own with the philosophies of Plato, Socrates, Aristotle, Kant, Descartes, Dewey. Hume, Marx and many others that challenged culture and laws.

If AI needs to learn those philosophies from the philosophers and can't devise a new philosophy on its own, it doesn't deserve the term intelligence. It's then just a parrotting copy cat.

The reason there are several cases of copyright infringement against AI companies in courts.

There are now well over thirty lawsuits that have been filed by copyright owners in U.S. federal court against AI companies, accusing them of direct copyright infringement for using copyrighted works without authorization to develop their AI models.

AI Copyright Infringement Cases: Insights From the Courts

How will the courts handle fair use decisions in pending #AI infringement cases? We take a look at orders that have been already issued.

copyrightalliance.org

 

[_TrXtR_]

Then it's not intelligent. Intelligence transcends laws and cultures by applying critical thinking. I seriously doubt AI could come up on it's own with the philosophies of Plato, Socrates, Aristotle, Kant, Descartes, Dewey. Hume, Marx and many others that challenged culture and laws.

If AI needs to learn those philosophies from the philosophers and can't devise a new philosophy on its own, it doesn't deserve the term intelligence. It's then just a parrotting copy cat.

The reason there are several cases of copyright infringement against AI companies in courts.

AI Copyright Infringement Cases: Insights From the Courts

How will the courts handle fair use decisions in pending #AI infringement cases? We take a look at orders that have been already issued.

copyrightalliance.org

I think we are a talking about LLM which would be one of the components of an AGI.
I think and AGI is what you are talking about. We aren't there.

 

[Cosmik Debris]

I think we are a talking about LLM which would be one of the components of an AGI.
I think and AGI is what you are talking about. We aren't there.

Can you write out the acronyms? Not all of us are in the IT world.

If the I in AGI means intelligence and we aren't there yet, why are we calling AI artificial Intelligence when it's just machine learning?

 

[_TrXtR_]

Can you write out the acronyms? Not all of us are in the IT world.

If the I in AGI means intelligence and we aren't there yet, why are we calling AI artificial Intelligence when it's just machine learning?

Because it's a buzzword.
Remember PC games have AI as well. Reaper bots can't do anything socretes did but it can make you cry.
And AGI will turn into a buzzword as well.
Artificial general intelligence.
But right now we have an LLM. Large language model. It is able to form a 'completion' of something with trained information. The more it's trained the more it looks smart. It's not. AI doesn't think doesn't known it doesn't feel it doesn't form an opinion, it makes a completion based on given text and trained info. It's just an advanced piece of code.
AGI is supposed to be where you get Will Smith movie type stuff. But that'll be dumbed down when they have trouble reaching it.

 

[Cosmik Debris]

Because it's a buzzword.
Remember PC games have AI as well. Reaper bots can't do anything socretes did but it can make you cry.
And AGI will turn into a buzzword as well.
Artificial general intelligence.
But right now we have an LLM. Large language model. It is able to form a 'completion' of something with trained information. The more it's trained the more it looks smart. It's not. AI doesn't think doesn't known it doesn't feel it doesn't form an opinion, it makes a completion based on given text and trained info. It's just an advanced piece of code.
AGI is supposed to be where you get Will Smith movie type stuff. But that'll be dumbed down when they have trouble reaching it.

Thanks. So my thoughts that AI is just machine learning and a copyright violation are correct. And the thread title is correct too.

 

[_TrXtR_]

Thanks. So my thoughts that AI is just machine learning and a copyright violation are correct. And the thread title is correct too.

Yeah. LLM is great to help solve baed in existing information. But if you say the wrong things you get rhe wrong responses, you also have to avoid hallucinating. So if you're coding, it's fine to generate allot of code bit you have to line by line verify everything. Same woth any work you do. So AI is not replacing tons of jobs soon. Just going to be companies that close their doors soon.

 

[konfab]

I think we are a talking about LLM which would be one of the components of an AGI.
I think and AGI is what you are talking about. We aren't there.

I don't think they will be suitable for that in their current form. The way humans think through a problem is far more involved than simply thinking about the next word to say. We think about the next image, sound, movement.

And those cannot be encapsulated by single word. You need a higher dimensional data structure. Not impossible, because after all, it is just numerical representations of data. But it would likely raise memory requirements by about 6 orders of magnitude.