YOU can try an AI chatbot as GOOD as LaMDA – Google’s “sentient” AI

Hello YouTube viewers, foreign, wide, welcome back to the Matt Video Productions channel. Now I am absolutely positive that you have heard of Google’s Alamda and the Google employee that was recently put on leave because he thought that it was sentient or alive. And I know you’ve seen many videos talking about it, but I actually have something new to bring it to this conversation. However, before we dive into that stuff, you guys know I like to open up these videos with an interesting AI generation. And this one I found particularly interesting for a few reasons. First of all, this is a silly and lighthearted one, and I have a soft spot for those. However, this is two gorillas arguing over who gets the last banana. And there’s a subtlety here that is kind of hard to, I guess, put into art, if that makes any sense. And that is they have to take gorillas that are typically not really capable of expressing facial expressions as well as humans are, and they have to convey that they are arguing over who gets the last banana. So the banana is definitely there, and that’s the easy part, but he’s pointing his finger and he’s arguing with him. And I just thought that was a really nice touch. I bet all of the variations of this didn’t have that sort of aspect to it where you can tell they’re arguing. But I just thought this was a particularly good display of the technology here. Not something that’s necessarily easy to do. I’d also like to mention thank you for the continued support to all of you amazing viewers who watch my videos. It means the world to me. At any rate, Google Lambda. Like I said, I have something new to bring to the table regarding this conversation. The cliff notes of the matter is basically a Google announced or is working on this Lambda conversation technology. This was talked about over a year ago in May 2021. And basically it’s an AI chatbot that can create convincing human conversation. And it’s also a language model. I’m sure it can do a lot more than just chatbot type deal. We’ll get more into that later, but a Google employee was basically testing it out. He thought it was alive based on the conversation they were having, the AI and him. And then he publicly released the conversation that he had with the AI, which of course was against Google’s policies. And he was put on a temporary leave. And then a bunch of news articles, of course, released. We’re just scrolling through all of them right here. Tons of people have been talking to this. I am not new to the conversation. And the general consensus is, of course, it’s not alive. If you know anything about this stuff, there’s tons of videos proving that it’s not alive. Really deeply explaining it. And while I might not be the person that can deeply explain it, you, the viewer at home, might as well already have access to Google’s Lambda. Because all of you at home can access OpenAI’s GPT-3 language model, which certainly rivals Google Lambda. We don’t really know what Google Lambda is necessarily fully capable of, because there’s no public access. But there is public access to GPT-3. And we know GPT-3 can pass the Turing test. Basically, the original test to determine whether something is, quote, unquote, alive. Obviously, it doesn’t really prove that this thing has consciousness. But we’ve had GPT-3 for a long time now. And I’m going to be honest, when I first used GPT-3, I was a little bit convinced that who was kind of alive. I didn’t really know much about AI stuff back then. So using it, I was, you know, it’s pretty blown away. It’s very incredible. And like I said, you guys can actually use this model. I will link it down below. All you have to do is sign up for an OpenAI account. By the way, these are the same people that make a dolly too. And then you will have a free trial where you can use GPT-3. And you get quite a lot of generations with GPT-3 after you sign up. And then you sort of pay per prompt or pay per token. After that, which I do pay for, and it’s very, very cheap. So to sort of convey this idea, the prompt is, the following is a conversation with an AI assistant. The assistant is helpful, creative, clever, and very friendly. And this is basically how OpenAI wants to start off the chatbot. So the AI knows that this is the conversation we’re having. And then it says, human, hello, who are you? And then it says, this is who I am. And then it lets us basically put in whatever we want. Now, keep in mind, I can delete all of this and put in whatever I want. I could say the chatbot is evil and wants to take over the world or that the chatbot is a walrus. And the AI will happily abide. And a lot of people have talked about this in Lambda videos, you know, explaining this is why it’s not alive, so to speak. But you can actually try this sort of thing out yourself with OpenAI’s GPT-3. You can talk to this thing as random as you want. I said, I have a large watermelon. There’s a hole inside of it. Should I look inside? Why is there a hole in it? I’ll submit that and we’ll see what it gives us. There could be a number of reasons why there’s a hole in your watermelon. It could be something that has eaten away inside the watermelon, which is, you know, something a normal person would extrapolate, something ate into the watermelon. Or it could have been damaged when it was growing. And if you’re curious, you can cut the watermelon open and take a look inside. Otherwise, there’s no need to worry about it. You can tell how human this thing sounds. It’s because it’s trained on all of that human data. It’s not trained on AI data, right? It’s trained on human data. And that’s exactly why it sounds just like a human. You can even ask you things like, do you have emotions? The AI can really decide whatever it wants with this question because it’s never explicitly mentioned that the AI has emotions other than being creative and clever and friendly. But it knows that it’s an AI. So it’s just going to say, no, I don’t have emotions. Now, if we go up here and say, the AI is very emotional, then it’s going to respond differently. As you can see, the AI this time responded, yes, I do have emotions. I’m very emotional, in fact. I enjoy feeling happy, excited, and loved. And there you go. There is plenty of training data in this AI’s database to express emotions and make them sound very human. And this is why Google Lambda tricked that Google employee. That shouldn’t even really be that surprising that it happened in the first place because many people that I’ve talked to have actually been tricked by this very AI right here that you guys can try. You can really change up this problem to be whatever you want. I said the following is a conversation with a monkey. He does not know how to speak and only responds with, ooh, uh-uh. And there you go. With just following instructions at the end of the day. And since GPT-3 is public access and a lot of people can use it, there’s a lot of interesting things that people do with GPT-3. I’m actually going to be making a video about a chatbond that you can talk to with your voice that uses GPT-3. And that is going to be an exciting video. You can also say things like explain the process of using a toilet as if it were a monster encounter in a Dungeons & Dragons campaign. If you’re wondering, this is from Reddit. The first step is to identify the toilet. It is a large porcelain bowl sent into the ground with a hole in the center. There’s a handle on the side which is used to flush the toilet. The next step is to approach the toilet. It can be done cautiously or with confidence. If done cautiously make a wisdom perception. Check to see if you notice anything unusual about the toilet. If you fail this check, you may still approach the toilet. But do so at your own peril. I’ve never played Dungeons & Dragons, but this reads pretty hilariously. Once you are within range of the toilet, make a dexterity, or a battox, check to see if you can safely climb onto it. If you fail this check, you fall into the toilet and take some points of damage. If you succeed the check, you are now standing on top of the toilet. From here, you must make a successful strength check to lower yourself onto the seat. If you fail this check, you fail. You fall into the toilet and take out over much damage. Once you are seated on the toilet, make a constitution check to see if you can hold your breath long enough to do your business. If you fail this check, you begin to suffocate. Oh my gosh. And if you succeed this check, you can do your business and fees. When you’re finished, make your strengths as a stand-up on the toilet. If you fail, you fall back into the toilet. So my point in bringing this up though is that this AI is capable of so many different millions of random little things, and it’s very good at it. This is a creative thing that it can do. That’s sort of the whole point of this artificial intelligence is that it’s artificial intelligence. It can come up with stuff crazy like this. It can come up with pretty much anything you imagine. It could write papers. It could write short stories, high-cuse poems. It could sound like a monkey or a two-year-old. Whatever you really can imagine, it can kind of do. I mean, there are limitations, of course, but it’s very unlimited, so to speak. And that’s why, you know, this stuff is getting a lot of traction and popularity. And why some people think it’s alive. And as you guys can see, with the quick line of tanks, it can even come up with a decent outro to an AI video. So there you have it, a brief overview of AI. As you can see, AI is fascinating and a complex topic with a lot of potential for both good and bad. As AI technology continues to develop, it’s important to stay informed about the latest advancements and to think critically about how AI will impact our world. Thank you very much for watching, folks. And I will see you in the next video. Goodbye.