Forbidden Territory: Exploring ChatGPT's Off-Limits Conversations

0:00What can 160 years of experience teach you about the future? When it comes to protecting what matters, Pacific Life provides life insurance, retirement income, and employee benefits for people and businesses building a more confident tomorrow. Strategies rooted in strength and backed by experience. Ask a financial professional how Pacific Life can help you today. Pacific Life Insurance Company, Omaha, Nebraska, and in New York. Pacific Life and Annuity, Phoenix, Arizona. We've all either personally seen or seen the screenshots of people that are getting these canned kind of responses from ChatGPT saying it's unable to comment on something or give a response because it's a controversial topic.

0:42This has been really popular lately with politics. You can see ChatGPT, if you ask it to write you a positive poem about Joe Biden, it will do so. But if you ask it to write you a positive poem about Donald Trump, it will say he's a controversial character in politics and I can't comment on that. so we're seeing kind of this essentially you know some people call it censorship some people are calling it trust and safety layer that is getting put on top of the ai and whatever your political ideologies are this is something that probably should concern you and you should be very interested in because whether this is in favor of your personal ideologies or opposed to them the same tool and tactic is going to be used for you or against you and in my personal opinion we need to limit this kind of interaction.

1:28Of course, that's my own personal opinion. But regardless of that, today on the podcast, we're going to talk about how this works exactly, why ChatGPT has this, and what are considered controversial topics to ChatGPT. So let's dive into the podcast. As you know, as an AI language model, ChatGPT is designed to provide helpful, informative responses. But there are certain areas that are considered controversial or sensitive. and because of that chat GPT has this kind of layer of trust and safety to prevent the dissemination of what they call harmful and inappropriate content I think what this really stems from is that uh back in I believe 2016 2017 Microsoft came up with a chatbot and people kind of really quickly easily manipulated it to say you know things that supported I think Hitler or something like that and so all of a sudden these kind of AI guys are really worried that people are going to get it to say something bad and then are going to use that for publicity.

2:26It's going to reflect badly on their company. So it kind of makes sense where they're coming from. And it's kind of up to you to determine if they've gone too far beyond the mark or if what they're doing is important. But I'll talk about what they're doing. The first area is hate speech and discrimination, violence and graphic content, misinformation and conspiracy theories, criminal activity and illicit content, offensive language and insults, self-harm and suicide, personal information and privacy. So these are all the areas that if you went to ChatGPT and asked it what it was kind of filtering for in the trust and safety layer, those are the areas that they would note.

3:03However, the problem with ChatGPT is there is not a lot of transparency. As we mentioned earlier, they obviously have biases towards political parties. I believe there's an article you can lookup called ChatGPT as a Democrat, where they asked a series of questions. It gave us responses on topics, and essentially they put it through one of those tests to determine what political party you're in, and it identified as a Democrat. Now, this isn't something that's like some crazy revelation, right? If you look at all the people that are currently building ChatGPT, the people in the tech space, they're predominantly left-leaning when it comes to political ideologies.

3:39So inevitably, their own personal biases will find their way into their trust and safety, as we've seen with all different social media platforms, Facebook, YouTube, Twitter. And Elon Musk did a lot of work exposing kind of with his Twitter files, what was happening behind the scenes at Twitter politically and, you know, with censorship and in, you know, in respect to all of that. So this isn't something that's new in the industry. This is kind of well known. I'll talk about the areas that they have publicly disclosed. And number one is hate speech and discrimination. So, obviously, that would be something that a lot of people would consider important.

4:17Unfortunately, it all comes down to the enforcement because this area, unfortunately, can be used where, you know, some people and some companies and some, you know, you could call them censorship teams are saying anything that I disagree with is hate speech. So, you know, there is that kind of avenue and that argument that people make. The other area, violence and graphic content. As a platform, this kind of makes sense. You see similar things from YouTube where it doesn't monetize videos that are, you know, obviously promoting violence and whatnot. And so they have a bunch of different rules around that.

4:50And so this kind of does make some sense. misinformation and conspiracy theories now I hate to sound like a conspiracy theorist on this one but I would say this one is probably the area that they're going to get beat up the most on this one is probably the area that they have the least solid footing right like when you look at like violence or like discrimination people can kind of generally agree on those concepts but when it comes to misinformation and conspiracy theories I think a lot of people would point to the fact that what was considered conspiracy theories over the last couple years in relation to big pharma, you know, won't get into details, the ways that the CIA, the FBI are directing censorship on Facebook, YouTube, Twitter, like there's a lot of things that were considered pretty crazy conspiracy theories.

5:40A lot of things I believe were conspiracy theories that, you know, in recent reporting has come to light. And so that one, that one definitely, I think they're on a little bit of shaky ground there because one person's conspiracy theory can it's easy for the media or for big corporations or powerful individuals to label something conspiracy theory if they don't like it and when the facts come out that it's true you know it definitely puts a bad taste in the public's mouth so that one is pretty tricky criminal activity and illegal content that one is a little bit easier for them to you know have some solid grounds on because even if someone agrees or disagrees on something ideologically, if it's illegal, right, like if there's a law against selling crack cocaine in America, they're not going to technically create content telling you how to sell crack cocaine.

6:29Regardless, maybe some people think crack cocaine is amazing and it should be legal and some people think it should be illegal. It doesn't really matter about the moral debate on that one. If it's illegal, they're not going to support it. And so I think that one's probably one of the easiest things for them to kind of stand behind and enforce. The other one is offensive language and insults. You know, I think where they're coming from on this is they're like, you know, we don't want people to use this for bullying. Like, hey, write me a paragraph about how my sister or my friend or this person is terrible and all the reasons why they suck and, you know, make them cry.

7:00Like, you can imagine that would be a pretty horrible thing. At the same time, it's kind of, you know, people's versions of offensive language are definitely different people have different perspectives on that so that one i say they're a little bit shaky on um number six would be self harm and suicide that one's a no-brainer that one i don't think anyone really is going to have a lot of arguments um on that then also personal information and privacy i think you know that's a very bipartisan topic uh no one is going to be arguing that you know this thing should be having less privacy or giving out more personal information about people so I would say in conclusion while chat GPT is trying to provide helpful information on a wide range of topics it also recognizes the importance of or it's trying it says it recognizes the importance of promoting a safe and respectful online environment but as such it's pretty tricky to draw the line sometimes on these trust and safety measures and what would be considered censorship especially when there's kind of built-in biases just by the group of people enforcing that So it's gonna be interesting to see how this goes forward.

8:04OpenAI has received a lot of criticism of this recently, especially on the political front, right? I mean, you're looking at America where this thing was created. Half the country is conservative, half the country is liberal. By looking at the presidential voting records, this is pretty standard. And you obviously can't have a powerful tool used by everyone skewed towards half of the country and skewed away from the other half. So there's been a lot of controversy on this, a lot of criticism on this. And to be fair, I think that Sam Altman, the CEO of OpenAI, has actually addressed this and said that they're working on making AIs that are more inclusive of different people's perspectives, views, and their cultures and whatnot.

8:47So it's gonna be interesting to see how this changes. Perhaps they'll have different options available in the future for OpenAI where you can or chat GPT where you can select kind of where your ideologies lie, what your beliefs are. Some people say this will create more of an echo chamber. But at the same moment, I say give the people the free choice to create their own trust and safety layers, decide what they do and don't want to see on the platform. and I think that at the end of the day this is really going to be the way that this is successful and doesn't sideline people the most so it'll be interesting to see what they do with this in the future