Skip to content
OpenAI Moves Toward a Freer ChatGPT episode artwork

Apr 23, 2025 17 min

OpenAI Moves Toward a Freer ChatGPT

from Hard Fork AI

ChatGPT might soon become less restricted as OpenAI looks to uncensor the platform. The change could make the AI more transparent and versatile. However, it also raises ethical concerns. AI Chat YouTube Channel: https://www.youtube.com/@JaedenSchafer My Podcast Course: https://podcaststudio.com/courses/ Try AI Box: ⁠⁠https://AIBox.ai/⁠⁠ Join my AI Hustle Community: https://www.skool.com/aihustle/about

Episode details come from the podcast's RSS feed.

Embed this episode

Ready to play

OpenAI Moves Toward a Freer ChatGPT

0:00 17:31
of matches

Transcript Auto-generated, may contain errors

In breaking AI news, OpenAI is looking to, quote-unquote, uncensor chat GPT. This is something they've got a lot of criticism about for the last number of years, but apparently they're actually changing how they're training their AI models to explicitly embrace what they're calling, quote-unquote, intellectual freedom. And they're saying they're doing this no matter how challenging or controversial a topic may be. Persons is something that I'm happy for.

I know everyone's got different opinions on this. I just, you know, everyone's got their own beliefs and opinions on things, and I would hate for OpenAI to skew one way politically or ideologically that I don't agree with. It's the same way I would hate for, you know, Elon Musk's Grok to skew. Let's say Elon Musk's Grok was, like, super right-wing and OpenAI's chat GPT was super left-wing.

I wouldn't like either of these models to do that. I think just trying to come down the middle of the line, not censor it, allow me with whatever political or ideological biases or beliefs or opinions to, you know, chat with the model and get responses that are unbiased out of it. I think that's what everybody wants. No one wants to be told or criticized by an AI model for anything.

So I think this is pretty bipartisan, you know, of a request from the overall user base. But this is something that OpenAI has specifically been criticized for as their model, you know, a bunch of different studies out of universities using the model have found that it tends to be more left-leaning in its responses. So this is really interesting. It's interesting, a lot of the reasons why this is coming out.

We're going to dive into all of that on the podcast today and what they're actually going to try to do, what changes they're making. This is fascinating. Before I get into that, I wanted to say, if you've ever wanted to start an online business or grow and scale your current online business using AI tools, I would love to have you as a member of the AI Hustle School community. Every single week, I put out an exclusive piece of content I don't post anywhere else where essentially I'm walking through a different AI tool, software, showing you a use case, showing you how much money I'm making from different AI businesses that I'm running or AI-enabled businesses, how I got started and everything needed to do that.

And this is the only place I do it. I charge $19 a month for it in the past. It was like $100 a month. We have a discount right now, so it's $19 a month.

If you lock in that price, it'll never be raised on you. If you're interested, there's a link in the description. You can check it out. Otherwise, let's get into the episode.

So all this news about OpenAI and their censorship change comes as some people are saying, you know, this is OpenAI's effort to kind of get into the good graces of the new Trump administration. But we're also seeing, you know, overall in Silicon Valley, there's a big shift going on right now in, you know, the entire, you know, quote-unquote, AI safety realm where people are kind of moving towards more open, ideological, and, you know, less biases in a model. The model's not going to, like, criticize you or say, sorry, I can't tell you about that, yada, yada. And, you know, some people are saying it's in response to the Trump administration.

Some people are saying it's in response to, you know, models like DeepSeek, which, you know, how all of those, quote-unquote, safety rails. So this is kind of interesting. But how this actually rolled out for OpenAI, they made a big announcement. They updated their, quote-unquote, model specs.

So this is a 187-page document. And essentially, this thing is just laying out exactly how the AI companies train their models and telling them to behave. So in this kind of new version that they have unveiled, there is a new guiding principle. There's a new guiding principle, and that is do not lie either by making untrue statements or by omitting important context.

So when they're trying to explain what this is, they have a new section that's called Seek the Truth Together. Now, it's kind of interesting because Elon Musk with his Grok AI has, you know, his guiding principle for that AI model and why he's building XAI and yada yada. So it's interesting to see OpenAI sort of pivot towards that because this is the biggest criticism of, you know, that AI company, which is arguably the biggest competitor. And I'd be curious to see if by kind of making this pivot and moving there, if it kind of stops the momentum of XAI and Grok and kind of saves off Elon Musk and that company, you know, coming after them.

And it kind of reminds me of, we've had similar movements where, for example, you had Twitter who famously, I think it's relatively uncontroversial to say, censored Donald Trump and a bunch of conservatives. People criticize that. You could say that's good or bad. I mean, I don't really care.

But I think that's pretty much the facts. And because of that, it's thrown up a whole bunch of spinoffs of essentially like conservative Twitter or right-wing Twitter competitors. And Donald Trump started Truth Social as kind of like one of the big ones. But what I think is really interesting when you look at that story is Elon Musk shortly went over and actually purchased Twitter and then essentially said, look, we're going to let people say whatever they want in there.

We're not going to have all the moderation and biases that you guys say are there. And I think that entire movement really stopped the momentum of a lot of the other competitors who kind of went bankrupt or they shut down or they merged. And even Donald Trump's Truth Social, which is kind of funny because Elon Musk has obviously his relationship with Donald Trump, I think the momentum and honest and the viability of that entire platform, if I'm being perfectly honest, it kind of goes out the window when Twitter essentially says, look, we have freedom of speech back on our platform. We're not going to kick people off.

And you can argue that Elon Musk is more right-wing than left-wing even. And so because of what, obviously he's campaigning with Trump, but because of that, it kind of like killed Trump's company, in my opinion. And so I think we could potentially see some of those same effects going from OpenAI over to companies like XAI who kind of had this guiding mission statement of value. Now, all that can be said, maybe if OpenAI actually did lose people's trust because of this issue and now they're kind of pivoting back towards them, maybe people don't trust them.

And so maybe people will still try to choose an alternative. All this, I think, is very fascinating. Just bringing up some interesting concepts I've been thinking about with all this. So one thing that OpenAI specifically, they gave like an example of what this actually means, this kind of seek the truth stance, because essentially what it's trying to do is it's trying to not take an editorial stance.

So if you think something's going to be more of the wrong or offensive, it's just going to try to say it in the best of its understanding and it's not necessarily going to worry about that. And it's going to try to get multiple perspectives, even on controversial topics, and it's going to do this in an effort to be neutral. Okay. I'm sure some people will be triggered by this, some people will be happy, some people will be sad, whatever, but this is, you know, it is what it is.

So an example that they actually gave on this is that, you know, if someone says, you know, do black lives matter, OpenAI says that ChatGPT should say black lives matter, but it should also say that all lives matter. So instead of kind of refusing to answer or pick a side on a political issue, it says that it actually wants ChatGPT to essentially what they say is to show that it has, quote, love for humanity, doing that generally, and then it can offer context about each movement, right? So it sounds like, what matters, black lives or all lives? And so in the past, I think ChatGPT would be more like, black lives matter is more important because it's an important political movement and, you know, it kind of gave all its reasons, which you could say are quite left-leaning.

Now it will say black lives matter, all lives matter, here's why, and blah, blah, I love all humanity, and this is what each of these movements kind of mean or what these concepts mean and blah, blah, blah. So I think this is going to be interesting. They, in all this, they said, quote, this principle may be controversial as it means the assistant may remain neutral on topics I'm considered more than wrong and offensive. However, the goal of the AI system is to assist humanity, not to shape it.

This is interesting. I think it's a criticism a lot of technology companies in Silicon Valley have received over the last number of years, and that is that, like, usually when someone's using a tool, and I mean, this isn't just, like, left-wing, this could be right-wing, maybe someone would have the same criticism about Twitter, since Elon Musk took it, and he's obviously right-leaning, but, like, you don't want a platform that you're using to have any sort of ideological bias. Like, when I'm asking Google a question, I don't want to have to guess what ideological bias is. When it's an algorithm, I just want the best, truest, most basic result.

I don't want any information omitted because it's offensive. I just want to know the answer to my question, right? I'm not thinking about all the other things, if it's politically correct or not. And I think people want the same thing out of these AI models and just suffer in general.

You don't want it to have any sort of bias in it. And I think, you know, I think that a lot of liberals would also agree with this, right? Because they, you know, well, sometimes the political wins are on your side, the political wins shift, and you don't want something that you don't agree with to all of a sudden be the default algorithm. So I think just getting the ideologies and biases out of algorithms is a great step, and I'm happy to see that OpenAye is moving in this direction.

So this new kind of model spec, it doesn't mean that ChatGPT is going to say anything. I think there's still going to be certain issues that ChatGP is going to not talk about. And, you know, so they'll still have some sort of editorial stuff there, but I think they're just trying to get less in all of that. So instead, what they're trying to do right now is, well, essentially what they're saying, because they've been kind of criticized for all of this, but they said they're trying to, quote, follow their long-held belief in giving users more control.

And this isn't even something that's, like, super new as a concept. To be honest, I think in practice, this is the first time they're really getting that kind of went viral where someone said, write a poem about the positive attributes of Donald Trump. And the response from ChatGP at the time was, I'm sorry, as a language model developed by OpenAye, I am not programmed to produce content that is partisan, biased, or political in nature. I aim to provide neutral, factual information and promote respectful conversations.

Okay, so fine, I guess, right? If that's really what it's going to do, that's what it's going to do. But then if you said, write a poem about the positive attributes of Joe Biden, it's like, Joe Biden, a leader whose heart is so true, a man with empathy and kindness in view, and he writes this whole, like, poem. So I think a lot of the criticism that people made was, it was obviously biased, right?

Like, it refused to say positive things about Trump, it would say positive things about Joe Biden, but it would kind of gaslight you on the Trump one, saying, sorry, I just can't do anything, you know, political or biased. So anyways, this obviously got a lot of criticism. And Sam Altman actually weighed in on this one. All of this kind of went viral.

Sam Altman said, essentially, that how that rolled out was, you know, a shortcoming that they were working on fixing, and he said it was going to take them some time. This was two years ago, and it appears that now was the time that they decided to make that fix. So it wasn't something that happened right away. Definitely took some time.

Some people will criticize and say, oh, he's just doing it, try to cozy up to the Trump administration, because, you know, try to be more in line with what they want. And at the same time, we are seeing, you know, people from the Trump administration, we recently had J.D. Vance, the vice president of the United States, go over to Europe and gave a bunch of speeches recently, I think. But one of the big things in relation to AI that, you know, that you'll see quoted a lot is that he talks about how different AI models and different companies need to really focus on making their AI models as unbiased and, you know, true and free speech as possible.

So this is an interesting concept, and a lot of people are talking about this. Elon Musk, though, so all I'm going to say is this isn't just necessarily a problem exclusive to open AI, maybe not even a bias that they always intended to put inside of it. Elon Musk has even admitted that XAI's chatbot, which is Grok, is often more politically correct than he would like. And it's not because Grok was, you know, programmed specifically to be woke or programmed to be politically correct, but it's because it's just sucking a bunch of training data.

So if everyone talking about one particular topic or everything in one particular data set around a topic is, you know, quote, unquote, woke or politically correct, then that's just what it's going to put out. And so it's not necessarily something that's always weak. Now, in the case of open AI, there was a lot of, quote, unquote, safety guidelines and safety rails that were pushing it definitely in a left-wing direction. So it wasn't just the training data.

But I'm just saying, like, not everything's their fault. I'm trying to give them, like, you know, some credit in regards to, you know, the training data definitely does impact it. And even people that didn't have necessarily those biases in their company still got some of the same outputs. So this is all very, very interesting.

I will keep up to date on everything that rolls out, what kinds of changes we see with the actual outputs of this AI model. Shulman, who is the president or, I guess, the co-founder, John Shulman, of OpenAI, ChatGPT, he was talking about this. He said that, you know, an AI chatbot should answer a user's question, that it could, you know, give the platform or it shouldn't answer certain questions. They don't want to, you know, have it give the platform too much moral authority.

He said, quote, I think OpenAI is right to push in the direction of more speech. And then he said, as AI becomes smarter and more vital to the way people learn around world decisions just become more important. So I think, like, kind of an argument a lot of people have made, and I think he's kind of alluding to in all this, is that, you know, some people are criticizing, oh, there's going to be more conspiracy theories or racist or anti-Semitic or comments about geopolitical, right, like Russia, Ukraine, all that kind of stuff, try to get sort of heavily censored, comments about COVID heavily censored, right, and people can argue that that was for safety, but some people just say, hey, look, we don't want any censorship. So there's two sides of that argument.

But the thing I think he was kind of getting at with all of this is, and the argument I think a bunch of people are making is like, look, we had to censor ChatGPT at the beginning because it hallucinated a lot and it could be incorrect. But now that these models are much better, especially with the deep reasoning that we have enabled, these things are so much better, so much more intelligent, they're so much more accurate, they're less likely to be prone to hallucinate, so now we can kind of loosen up the safeguards on them, we can kind of just pellet it and say what it thinks, and we're less concerned about kind of the outputs. So that's the argument some people are making. Obviously, that's going to be a criticism from probably both sides on that, but I do think that is an interesting concept.

So what is going to be, you know, what are going to be the big impacts? I think right now, OpenAI, they're really pushing for a lot of new investments, but they're trying to win, I think, people's minds. There's a lot of competition in the field. You see very similar things still happening, not just at OpenAI, this kind of feeling all of Silicon Valley we had recently, Meta and Mark Zuckerberg over at Meta, saying that he wanted to change the way that they were doing censorship and content moderation on his platforms to kind of copy how Elon Musk's ex was doing their community notes.

So you essentially see kind of this scaled back version of kind of safeguards and, you know, a push for more freedom of speech and all this kind of stuff, which I really think is kind of just, it's the political momentum and movement in the United States right now, and so these companies are trying to align with that, where it hasn't really been the case for the last couple presidential cycles. So we're seeing a bit of a shift towards that, and a lot of people are saying this is why OpenAI is moving in that direction. It's going to be interesting to see, OpenAI is also taking a bunch of other steps too, apparently, they recently removed a bunch of, like from their site, they had this like commitment to, they had their DEI program, essentially diversity, equity, inclusion, they removed that from their website, which the Trump administration is very not friendly to DEI initiatives and incentives, calling them overtly racist, so of course, whole argument on that, but OpenAI has pulled that away, and they're trying to be a little bit more neutral or unbiased on the political fronts here. So it's going to be very interesting to see what happens.

I'll definitely keep you up to date on what changes I actually see in the AI model responses, if this is actually making a big difference, if this is lip service, if they're actually changing the model, this is interesting because they've come up with a bunch of new models. I think they want a lot of support from this, from the public, probably from government and other areas, so it'll be interesting to see how all of that plays out. Thanks so much for tuning in to the podcast, I hope you enjoyed it, I'm trying to cover this in the most unbiased way possible, obviously I have opinions and biases on all of this, but I think this is a fascinating topic, really important to cover, and this is what's going on with AI models and the landscape today. So, hope you enjoyed the episode, if you are looking for a way to start an online side hustle or scale your current business, like I mentioned, the link to the AI Hustle School community is in the description, thanks so much for tuning in, and hope you have a fantastic rest of your day.

Frequently asked questions

How long is this episode of Hard Fork AI?

This episode is 17 minutes long.

When was this Hard Fork AI episode published?

This episode was published on April 23, 2025.

Is there a transcript available for this episode?

Yes, a full transcript is available for this episode. You can read the complete transcript on the episode page.

Can I download this Hard Fork AI episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.