Humanoid Horizons: Exploring the Future of AI in Robotic Beings with Stable Diffusion Drawing Abilities

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. In the past, I've talked a lot about the implications of humanoid robots and what that would look like if we are integrating AI into them.

0:36So today I want to talk about a really interesting story by a company that has recently integrated not only GPT into a humanoid robot, but also has integrated stability AI and essentially created the ability for this robot to draw, which is really, really interesting. So I guess the basic overview on this is the fact that back in 2021, Emeka was essentially an advanced humanoid robot that was developed. And it was given, it has a bunch of different equipment on it. So it has microphones, cameras, facial recognition software, and GPT embedded into it. So all of this comes from a company called Engineered Arts.

1:16And you may have seen them, I think about nine months ago, they had a YouTube video go viral where essentially they had one of their engineers interviewing this humanoid robot and, you know, talking about the applications of humanoid robots and AI and what it can do. And, you know, we've talked a lot about that on the podcast. I think this has had some really amazing possibilities for, you know, robots to take over things like electrical and plumbing and all sorts of very what were previously, you know, blue-collar jobs. and of course I don't say that in a positive way like oh the robots are going to come and take over all our jobs I'm just saying this like this is inevitably going to happen these AIs and these robots will be trained to do tasks right now when we see this AI revolution we primarily are seeing this in a way that you know we're having these robots take a lot of white collar office jobs or essentially automate the AI can automate a lot of office jobs and administrative tasks and businesses and I think the next step is when these humanoid robots come out the AI will then be able to start automating or doing a lot of blue collar tasks now this has pros and cons and you know to to take this from the opposite side because of course it's quite obvious the side of we don't you know plumbers and electricians are being very upset about this if all the robots come and take their jobs but from I'll take I'll steel man the case for the other side which is that you You know, I have a family member that is an AC repairman and has a very, very difficult time hiring people that are reliable to help with his business.

2:47And so he is, you know, he would love to retire, but he just can't quite do it and he can't find someone to replace him in his business. So he is forced to continue working, which is quite difficult. if there was a humanoid robot that he could train that could do all of his tasks for him I mean despite the fact that the people that he works with might think it's weird and all of the other side effects but like from a practical perspective this would be a lifesaver for him right if he could buy one of these and have this essentially run his business maybe he runs the phone calls manages it tells it what to do I could see a future where this is going to be very useful And in my opinion, this is like it or hate it.

3:25This is an inevitable future that will arrive. These humanoid robots will be able to automate these tasks. So beyond all of that background, what I want to talk about today is the fact that this new robot, it's called Ameka. It was recently upgraded with Stable Diffusion, which is a deep learning model for generating images based on text prompts. And it now can draw. So they recently did a demo on YouTube where essentially they asked it to draw a cat. And what's interesting is it explains a little bit about how it goes about this process. So the robot, which has GPT-3 in it, I really think they should upgrade it to GPT-4 now that the API is in there because we know that GPT-3 is definitely a less advanced model.

4:05I actually think this thing could reason better, talk better, do a lot of things better, literally just with the upgrade of the software API. So that's kind of cool that their overall product gets significantly better once OpenAI or anyone else releases a little bit better software that they can just upgrade. But in any case, it explains what it was doing and how it kind of integrates. And it said, I generate my drawing image through the open source neural network project Stable Diffusion. From there, trajectories of the drawing are available. And when I skeletonize the image and vectorize it, after that, I plan and execute the trajectory to draw the image on my canvas.

4:43So, essentially, the engineer that was with, I don't know if I said her because I was, in any case, there was an engineer there that told the robot what to draw and it drew a cat. Now, I'm not going to lie, this is not an incredible picture of a cat. Like, it is the right shape, roughly, of a cat with ears and the eyes, which are kind of funny and deformed. It literally had a dry erase marker and is drawing this on a dry erase board. so it's not very fancy it's probably something similar to if I gave my four-year-old a dry erase marker and told him to draw a cap maybe worse to be honest then the robot did sign its name at the end which I thought was pretty funny now the engineer that was with the robot that was with Emeka told it that when when the robot essentially asked you know if if there was uh how it was or to rate it or whatever and the researchers said that it looked very it said it's kind of sketchy and the robot said if you don't like my art you probably just don't understand art which I thought was pretty funny I'm not sure if this is some sort of canned or programmed response or if it literally is saying that in any case it's pretty funny but I do see a future where this concept I think is really powerful of not just like you know we've seen all the automations before where you have like an automated robot that essentially just moves its hand to the perfect place and can color something perfect or you know we could probably train a robot to be a printer and just do everything exactly perfect but i think this is an interesting point where they're training robots to do this in a different way they're training the same way humans do it by using cameras and spatial recognition software and all sorts of different things that are not necessarily just like a rigid printer per se and it's able to go and actually learn how to draw and and draw things and i think integrating with stable diffusion is really interesting because you can imagine stable diffusion or a lot of other different image generating software like if you had something like mid journey integrated into one of these robots and then had it you know paint something that it was you know generate some sort of beautiful image in mid journey and then it's trained to paint and it could paint that like these robots could create all sorts of really incredible pieces of art um really incredible things so i think this is a really interesting application obviously this goes beyond that you know it can draw blueprints and all sorts of different things that it could probably come up with that we can train these ais to do but i thought this is a very interesting use case um and a big you know it seems like a small thing a lot of haters or a lot of people will be like oh look you know it drew a terrible picture of a cat but you know i think it wasn't that long ago when we're saying, oh, look, this GPT, you know, had a bunch of grammar errors and like said something fake.

7:29And then all of a sudden it gets better and better and better. And all of a sudden we have GPT-4, which, you know, admittedly is not perfect, but is quite complex. I think it's gonna be the same thing with art, where these robots, you know, the first version is a kind of a gimpy looking cat picture that people are like, oh, look, it's not that good. Haha, it's terrible. It's like, okay, well, what is it going to do in a year? What's it going to do in two years? What's rate of AI, this thing can paint a masterpiece. Honestly, like imagine you have a wall in your house and you could get one of these AI robots hooked up to mid journey to come into your house with a bunch of paintbrushes and go and paint an incredible mural on your wall.

8:06Like this is inevitably going to happen. And this is going to be literal paint. In fact, imagine a point when artists use this to paint things for them. I'm not saying this is good or bad. Actually, I mean, obviously it's bad but like imagine an artist getting a robot to paint something for them and then passing it off as their own work i mean like no no this was mine like right now we're talking about what are we going to do to have like ai detection in text like how are we going to text detect if this is written by ai okay here's a bigger question how are we going to detect when something was painted by ai when a creative asset was created there's so many more um interesting things and then like it gets to the point where let's say we have like ai electricians in the future or like these humanoid robots that are also have ai into them and they could be trained to be an electrician let's say one of those goes into a home wires something wrong but the owners weren't there um the owners come back let's say there's some sort of liability the house burns down then the electrician says oh no i i did that myself it's you know registered under i'm bonded and insured this should go on off on my insurance but really it was like the robot like how do we prove like without cameras or a ring camera or something else like assuming that's not there right?

9:13Like, how do we prove that it was the human or the person? So I think that this whole AI detection is going to, it's going to come in ways that we're not thinking of right now. Right now, we're just thinking of like, how do we detect AI text? Okay, but how do we detect the handiwork of AI or artwork of AI? There's gonna be so many different areas that people are going to be looking at and building into in the future. So very interesting space and something I think we definitely need to continue watching.