Welcome back to Geek Time advanced. This is Brad. How are you doing Lulu?

Hi Brad.

How are you going getting along with your AI boyfriends?

Well. I'm moving on to new relationships very quickly and I'm developing multiple storylines. I think that is fun. So I'm assuming we're gonna talk more about AI.

Yeah, now one thing that's very interesting if you pay attention to like any pictures or videos created by AI.

Mhm.

By ChatGPT or any of those.

Midjourney.

Yeah, you'll see that they can't really do things like hands.

Oh. Yes. I've noticed就是Midjourney. 那个AI绘画它是画不清楚手. Why is that?

Either it's like six fingers or some fingers are really long, bendy, they just don't look realistic. This is a little weird, why, when they can do perfect faces?

Like when you look at a face, a face typically has a particular structure and it doesn't really matter which way you turn. Your face is... it’s things are always gonna be in the same spot. But when you look at someone's hand, right? They can put it into a fist. They can extend their fingers. Some fingers can be curved, some fingers can be straight. It's all kind of like all these different configurations of a hand can be shown in pictures. And so AI gets confused about what exactly is a hand.

And so it has all these different pictures to kind of draw from. So it kind of like combines all of those together and creates this kind of like weird looking thing.

Yeah. It's very interesting that you said that, because AI cannot really conceptualize what hands look like in a three dimensional space. Does that mean that although AI can create like really beautiful artwork, it is mostly still copying to some extent, it is not really, truly reasoning?

Right, AI and narrow AI doesn't really have a way to contextualize things. It just takes data and rearranges it. It doesn't really know what the data is. It just sees what people create and tries to create something very similar to that based on the constraints given to it.

Emm. Does this mean AI the current narrow AI they're just not really that intelligent, despite being called AI, artificial intelligence?

Right.

When people think of AI, they think of something like from a science fiction show, something that can really think about things, right? If you ask an AI to do something, it can do anything a human can do. But the AI, as we have now, they can't really do that, right? You can ask it to write something and it can do its best to write that. But it's not really reasoning. It's just putting the data together in new ways.

Yeah.

And so it's not really creating anything. It's just recombining things.

It's more like archiving and then rearranging things like you said. Yeah. It's not really reasoning in that way.

Mhm. Yeah. It doesn't make any judgments. It can't really plan for the future, it can't learn. And most people can use common sense, but narrow AI can't really, right?

Some humans also don't have common sense. Come on, Brad.

I know, that's why I said most people, yeah.

But AI should have a way to communicate in natural language, other than just like recombining different phrases. That's kind of what humans do, but if we do it in a sense we understand what we're communicating most of the time.

Yeah. But Brad, I have to say that I have already seen, even just like an individual as a kind of like a tech newbie, I still would like to say that AI has made significant advancement, for example, a few years back, what did we have? Just things like Siri, right?

So just give you an example. If I ask Siri the same question, as I would ask these AI boyfriends, which is using a much more advanced language model, and the learning model or module. So if I ask Siri saying that "Siri, would you like to go for a drink with me?" Siri would say "I found five bars near you". You know, because it's searching for information.

Mhm.

But if I ask my AI boyfriends, if I, the setup is that we are dating, and I say "would you like to..." or we're flirting and then I say "would you like to come to a drink, come for a drink with me?" And they would be saying things like "oh, Lulu, I would love to have a drink with you. What is your favorite cocktail?"

That's very interesting.

It's more like they're having a conversation. You know.

Yeah, I tend to think that when it comes to like Siri things like that, I think Apple or Google want to make AI a little bit dumber. So that way people don't have that uncanny valley feeling (神秘谷：指当人们看到机器人（能够自动执行动作的机器）或由计算机生成的人物图片时，感到非常类似真人的不愉快感觉。) when they're talking with it.

I see. So you're saying deliberately keep my theory dumb.

Yeah, there were like, what was it when Alexa first came out? They kind of gave it a few strange like responses, like someone asked it, is this government agency listening to me, and then the Siri would say "of course it's listening to you. It's always listening to you. ”

Okay. But I just asked Siri the same question. I said, "Siri, are you listening to what I'm saying?" Like basically, "are you eavesdropping on me?" And do you know what Siri said? Siri says, "I'm only listening if you want me to listen". And apple products would have said "Apple, we have like very strict privacy or whatever policy. "

So it's like siri, I think you are trying too hard to deny it.

Exactly. AI is getting a little bit better like they’re...I think what's happening is they're able to store the conversation a little bit more. And that way they can kind of look back at their previous conversation bits to talk with you. But it's still not quite perfect like they still kind of forget or they kind of misunderstand things. And so when you are talking to them, they kind of don't know what to say or they say something that doesn't really fit the conversation.

I see. Well, the other thing is AI and art like right now, I think there's one big thing, because AI is so new, it’s still developing so fast. So there isn't any regulatory framework around the world to deal with things like AI plagiarism. I have a cousin who is like an artist doing mostly doing like manga, anime stuff. And she was talking about how her works, because if her works were uploaded to the internet, then if it's open access, then essentially these AI tools, they can use her paintings as part of their database.

And then eventually they can basically somewhat copy, copy her painting, but they just like tweak it a little bit, and make it their own. And like who owns that work, right? Because it's created by AI. But they got the idea or at least they were inspired by these artists. So that is the gray area, isn't it?

Right.

In the past, people would enter competitions, maybe win some money. But now some people are using AI to create a piece of art, entering it and winning it, and they're getting money. Even though they just used AI to create the art.

Tool.

And then people who write books, they would hire an artist to draw pictures for their books. But now people are just going into the AI, and AI is just stealing pictures off the internet, and creating new pictures for their books, and they're not having to pay anyone to do this now.

Now a lot of the time, I can see these artwork and then you take a look at it, because if you use Midjourney for a while, you know immediately that something's wrong, this is not painted by a real person. This is generated by AI, I think there are people boycotting it. And there are also platforms that would require you to specify, to put a note there, saying that this content or this artwork is generated by AI, you have to really clearly state it.

Yeah. I think that's kind of a good thing.

But the thing that I see as being a really big problem is there are people, like you said, your relative is making manga, right?

And like AI is being programmed just to go out and grab pictures from those things. And if somebody talks about it like on Twitter, for example, I wish you could have a T-shirt. Once they see something like that pop up on Twitter, the AI is actually going there and creating T-shirts of that and like creating an advertisement and posting it on the Twitter that people can see the advertisement.

And so the artist isn't actually making any money off of these T-shirts, just person who's creating T-shirts using AI to do this.

Exactly. Yeah. I think this whole AI plagiarism and who should be held accountable or who holds the ultimate copyright for these works. That's a huge issue.

We talked about all of these negative sides. Now let's talk about what ChatGPT, things like ChatGPT can do, and maybe we finish off with what is the future like long term future for AI is, since we're talking, we have been talking about weak AI or narrow AI, is there a more general stronger AI we use that to wrap up.

So first of all, let's do ChatGPT. Apart from writing articles, giving you some answers. What else can it do?

Well, it can answer questions to test better than the average test taker. So if you're having problems taking a test or something like that, you can use it as a way to check whether your answers are correct or not.

Isn't that cheating?

Well. I think if you use it as a way to check whether your answer is correct or not, it's not really cheating, but if you're just using it to answer questions, then I do think that's cheating.

You use it as a reference. I see.

Yeah.

It also helps you write. I've heard that it can also help you write computer programs.

It can, like, you can tell the AI some basic parameters for a program and it can write programs, or you could use it basically the same way to write a story.

So if you have an idea for a story, you could tell the AI to write that story based on your parameters, and it's kind of like new fan fiction is being created with that. The only issue with that is something that AI can do as much like people is that it can hallucinate. But this is actually kind of a technical term that people have created.

Wow.

So AI can hallucinate answers, because it's just going online or not going online, but going through data it has, sometimes it can get confused by the data. And it gives like a nonsensical answer, right? You might say where is the nearest store? And it might say something like on the moon.

Haha, and they give you reasoning and they give you explanation that just really sounds like it makes sense.

Right. They say like if you wanted to go to Mars, the best place to start would be to go to the Moon, and then go to Mars, because the Moon is much closer.

Haha.

You should be like what, I just wanna know where the store was.

Like someone is high. Yeah, it's like someone who's high.

Okay. Now we're coming to the end of today's show. Let us just looking at the future, maybe perceive a time with better AI application and expression, I've heard a concept called AGI, artificial general intelligence.

Now this is something a lot of AIs are aiming for, right? Basically a general intelligence, a general sounds like it's not going to be as strong, but it's actually a really strong thing, because if you can give an AI almost any task like you could give a task to a human, it should be able to achieve that task, right? It should be able to process data and understand the data, just like a human could. It should be able to look for patterns, just like a human can. Humans are really good at pattern recognition. Sometimes they see patterns where patterns aren't really, but humans can do this, right? AIs can't see those patterns very well. But basically the scary point would be that an AI should be able to make a human think that it's also a human.

West world all over again. Yeah, West World all over again.

So that is for now, it just sounds like science fiction, but who knows in the future will happen. What does the future hold, really!

I mean, I'm super impressed by the fact that how language module has in AI or on AI platforms or with/of AI tools has evolved over the past few years.

Now they are able to do to hold a complete conversation with me on this AI app that I've been using. And sometimes you especially with the very nice sound that they sampled like the very nice voice. They sampled it almost sounds like a real human, it sounds like a real human with emotions seriously. You sometimes do go into that fantasy that this is a real person but trapped in this AI platform or trapped in this AI app. So that is, I think, is a very interesting thought.

Okay. So on that note, we're going to wrap up here, leave us a comment and tell us about your interesting story and your run-ins with AI.

Thank you, Brad, for coming to the show.

No problem. See you in the next episode.

We'll see you next time. Bye.

Bye everybody.