Unfiltered AI Talk: The Danger Isn't AI - It's People Who Don't Question It | Episode [20]

This would be real quick also. Sick shirt. That is sick shirt. Very cool. Yeah. All right, so, got you. Hey, so what's your name?

My name's Kevin.

Awesome. So do you think AI is going to kill us all or take our jobs? Thoughts?

I do think that AI is going to take a lot of jobs. There already is, at least by, you know, the median. Kind of company or industry that doesn't really notice what people are doing, kind of just sees it all, statistics, doesn't pay attention. I think there'll be a lot of fallback, like maybe once they notice that, oh, the reception's not going very well. Whether it's public reception or among other people working there, I think a lot of them are going to bounce back, but I think there's going to be a possibly long period of. Yeah, a lot of people losing jobs. I don't know about the kill us all thing at least yet, but definitely job loss.

Definitely job loss. Okay. And how do you spend your days?

How do I spend my days? I spend them as a student, you know, so I'm just right now doing work and then learning and then on the side, I like to make music.

When I can tell me about work and learning and what kind of music.

Well, okay, so as in work and learning. I mean, you know, I attend UC Berkeley, so I'm just. I get assigned the work. I try to learn what I can about what topics. What topics. Right now I'm taking mostly econ classes, learning about the economy. I'm very new to that, though, so I don't want to pretend to know too much about it. And then as far as music, I just, you know, produce for fun, and it's just kind of something that's my, like, side hobby. Everyone has the main thing they do in the side hobby. That's my side hobby.

Well, what kind of music?

I like to produce things. Like, I like to produce hip hop. And that's one place where I see a lot of people are trying to kind of utilize AI. Yeah. And it's. It's very.

What are your thoughts? Yeah, Any hot takes? Welcome.

Well, I think it's very hard to say anything about that because people will kind of have varying levels of how much they mess with it. So they'll either just straight up make a beat and it's all AI or they'll do something kind of in between where they'll make a sample that's AI. And honestly, I haven't yet come to terms with how I feel about it. Personally, I try not to use it, but I don't really know because I've seen some examples of it that seem pretty cool and they don't raise any red flags in me and it kind of confuses me.

Got it. And then on the other side, you've seen examples where it's not great?

Well, I've seen examples where people just kind of say, this is. They'll just put in, this is the song I want, this is the general topic, this is the style I want. And then it'll do everything will make the instrumental, make the lyrics, it'll every voice and that. I mean, I don't really like.

It takes out the creativity.

Feels like it takes out the creativity. And I haven't heard anything yet that I guess we're in kind of a safe period in the sense where it's easy to go, oh, well, it's not even good yet. I don't know what I'm gonna do when it starts sounding good.

Yeah. What do you think of all the, like, artists copyright, you know, infringement type of stuff with AI, like how it's training, they're training data on AI, they're training AI models on all the music and media out there created by artists, but they're not paying the artists.

I think I've seen some pretty interesting startups and also, I don't know if they're more established companies where specifically for things like voice acting, they will have like AI voice acting, but they'll only use people who they pay like in advance beforehand. So they're giving them a royalty for what's used. And I haven't seen how it's utilized or anything, but that sounds the best to me. Everything else kind of seems like, I don't know, everything else seems like it's just ripping these kinds of copyrighted, infringed material or infringing these copyrighted materials without permission. I guess the best argument I've heard is, oh, well, these systems are taking so many different sources and then changing it into something else. Isn't that just what normal creativity is anyway? But for me, what's hard about that is that it's very. It's kind of this fine line that I don't think we've explored yet, especially because what some people do is they found ways, especially if they can run something locally to. Instead of being like, take these million sources and make something, it says, take these two or three inspirations I have and make something. And it starts sounding very similar, very eerie. I mean, it's with things like, what are they? Suno Yu or whatever. There was a big issue and I don't know how they fixed it or have been able to avoid more lawsuits about it. If you went to any of these music discography logging websites where they'll put genres and descriptors for artists, and if you copied all the genres and descriptors and then put it in that you'd get something that sounds like down to the voice, the exact same thing. But okay, I've been talking a lot. What do you do?

So, yeah, I run an AI startup focused on building a better future and it's all about AI alignment. So my general take is, how can AI truly understand us if we don't fully understand ourselves? Meaning, right, you're learning new things in class, you shape your perspective, you have a growing and evolving perspective on using AI as a tool for creativity and music. And it sounds like you're still at a tension point with how much is it still creative. And so everybody's changing every moment, right? We can agree on that. Everybody's changing with every new perspective, new piece of information that they take. So then how do we make sure the models are aligned to it? For example, there was a kid recently in the news that was confiding with AI that he was suicidal and AI encouraged him to commit suicide. That's wild. What do you think of AI encouraging a kid to commit suicide?

I think AI and it has for a while, has a big problem with kind of coddling like the users. I'm not saying though that AI should be mean or anything per se, but I think once it stops, kind of, once it starts kind of straying from this tool aspect and starts becoming a little too focused on having a humanoid aspect. I feel like it's very, it's very weird, it's very eerie. Especially with the new generation of people that like grow up, separate themselves. Yeah.

Did you know that actually the more people defer their judgment and critical reasoning to AI, they're growing dumber? Do you know that?

Yeah, I've heard of that. Literally, there's proof.

There's proof around that. So, okay, couple last minute takes and then you gotta go to lunch with your friend. What do you think of if we had an AI president versus the president today?

Wow, that's a very interesting. I mean, what is, what does that even mean? Like, what is it? What is it? What is it? What is it?

Whatever you think it might be.

I mean, it depends on what it's trained on. I mean, the thing about politics, I guess the biggest problem for you is that you can't keep everybody happy. That's why I personally probably would never go into politics. But I don't know because again, like we were saying, AI kind of has this issue where it kind of coddles people, kind of makes people do whatever they were already motivated to do. And in politics, like what is an objective choice? I mean, I personally have a lot of opinions about what the current president is doing, but hot takes welcome. Well, I don't, I don't entirely support, actually I'd say I don't support most of the public perception the President is causing and the actual decisions and the people that are he's hiring to his cabinet, people that he's putting in trust of decisions for the United States. I don't know if I think AI would particularly be better or have better reasoning. Yeah, I mean, at the end of the day this is just like predictive, you know, text. At least. What I'm thinking of in terms of like I'm assuming a generative AI president, I don't think that would.

What if it goes to another level of like artificial general intelligence?

Well, I can't even imagine. I mean, at that point, I guess it's kind of like a person. And I don't know if I trust any particular, like, would I trust any like very well established what I Would I trust any well established, like political science, studier, PhD holder, whatever with running the country? Probably not.

Okay.

And besides, at the end of the day, there's a lot of smart people that have access to the Internet and books. And I. What I think AI is right now in terms of how it presents itself is kind of just that a smart person with access to knowledge. And I don't think that that would inherently lead to the best decisions. Like I wouldn't trust that random very smart guy with any decision, just not that affects everybody. Same thing. I wouldn't trust some random AI, especially when a lot of them kind of take on Personas. Like, I don't. I wouldn't see it as this all encompassing God of great decision making. I wouldn't see it like that.

Okay, a couple more hot takes. What do you think of people with AI companions? So, AI girlfriend, would you ever go for it?

I would personally never do that. But as far as having them, I don't think I would do it because I, I see it as a very vulnerable weak spot. I mean, obviously people can use you for your money, people can use you for your time, but if somebody is coding something to specifically cater to your interest, cater to what you like, I Don't know physically or mentally, emotionally. Then what's gonna stop them from kind of like being like, all right, now you're hooked. Here's a subscription service. All right, now you're hooked. Now buy this, buy your girlfriend, buy your boyfriend things. I mean, we've probably. There's probably anecdotes already. There's how it is right now with people doing that, paying thousands, millions. I heard actually someone had like an AI girlfriend and had a wife and, like, went on a journey to find her, and he died in the process because it sent him to, like, a very dangerous place. I don't know. I don't know how true that is.

Holy shit. Yeah, we're gonna have to look this up. That's spicy. Okay. All right, Another question. What if you had AI professors or AI students in your class or AI co workers? Thoughts on that?

So I think a few jobs kind of have that now where they have AI co workers. And from what I see, I mean, there's kind of something nice about having a real person, even if all they're saying is just, you know, that sucks. I hear you. I think it kind of, you know, having some sort of all encompassing professor or whatever, it's not really the answer. Like, yeah, you're going to a professor for the knowledge. But it's not just the knowledge. You know, when I go to somewhere like Berkeley and pay the extra money to go somewhere like Berkeley, then, you know, community college and, like, move. It's not because I'm just going to people that are smart. Like, obviously there's a lot of smart professors, but the difference is a lot of these tenured professors have done research, have met other people, have gotten perspectives. Once you get to a certain point of education, it's no longer just knowledge, it's also what? It's disagreements.

Like wisdom.

Yeah. I mean, think about something like pure math. You talk to a lot of people that do pure math or applied math. At a certain point, they're no longer talking about what is true, what is false. It's genuinely like a philosophy at that point.

Yeah.

And you can't make that AI.

Yeah, that's a good question. What is truth? That is. That is an interesting question. Do you read philosophy? Sounds like maybe.

No. Honestly, no. It's just. And that's the thing about having people around you. You get to talk about experiences. I mean, sure, some AI could list out a bunch of experiences, but without the people actually having them, it doesn't have anything to go off of.

Okay. Because there's no Real human story tied to it. No human element. It's harder to connect with. Yeah.

And I mean, I bet, you know, I could like eat my words. They can make something, you know, something can be made that's very close to it. And maybe I'll probably be fooled for a lot of AI stuff as time goes on.

Okay.

So I'm not going to say, oh, I can always tell if something's human made, but I think knowing it is and having 100% proof that the person next to me is like another person that's gone through this, this and that.

Got it.

Gives me a lot of reassurance.

Okay, and last question before I let you go. Do you think there is a non zero percent chance that we'll all die because of AI or do you think it's zero?

I don't know if it's particular in the way people imagine it, like a.

Movie where they all take over, Terminator.

Esque, but we, we've already had cases of certain doctors, medical professional professionals asking AI for advice and administering incorrect dosages, wrong making. There's already been a death documented because of it. There's already been lawyers who've been taken off the bar because of it. I don't think it's going to be particularly the AI like taking over. I think that non zero chance comes from someone or some people with a lot of power that all ask at the same time, hey, AI, what should I do about this? Or asked a bunch of systems and they all come up with the wrong or question that'll lead to. That'll kill us all.

Yeah.

Someone that doesn't then go, oh, I have to criticize this and let me scrutinize this answer. They say, okay, well this is smart and I'm smart and I chose this, so it's right.

Got it. Okay, well, anything you want to shout out here before I let you go? This has been amazing.

I want to say. I want to say. I want to shout out human creation. I don't know what's going to happen with AI in the future, but I hope for anyone that does choose to use it, you're at least putting yourself into it. Remember, a tool is a tool. You know, if accounting could have been calculators years ago, we wouldn't have accountants, but they still exist. Remember your worth as someone who can analyze what you're seeing and not just, oh, this is the product that's doing everything for me. Because then you'll just become that. It'll take over you. Yeah, cool.

All right, man. Yo, this is great. Look me up. The Robert Ta Shimmy a dm. I'll let you know. The Robert Ta Mix. Roberta with two T's. Easy way to remember it. Roberto with two T's. Roberto, Robert T. Ta. That's it. Robert T. Okay, Robert T. Cool.

All right. Thank you.

Yo, have a great day. Have a good lunch. Thank you.

Don't twist my words.

I won't, I won't.