Skild AI raises $1.4B at $14B valuation to build a general-purpose robot brain for any hardware

How are you doing? Good to have you on the stream. Thanks for

Great setup you got here.

Fantastic setup. Thanks so much for

testing it for the first time and as you can [laughter] see there a few issues in between so we do not know.

Yes. Uh please uh introduce yourself and the company since it's your first time on the show.

Yeah. So thanks for having me. I am Deepak Deepak. I'm a CEO and co-founder of Skilled AI. So at skilled what we are doing is we are building a general purpose robot brain. Now

any robots any task one brain. So when I say that it's like even a humanoid or quadro robotic arm hands everything governed by a single brain and that is what we do.

Uh let's get into the news from this week and then we have a bunch of questions.

Sounds good.

Uh what what happened this week?

Well there are this has been a busy week for us. uh I think back to back we had multiple announcement. Uh we began with uh the the announcement regarding learning from human videos. We showed a lot of results but the latest announcement which is just from uh yesterday is about the funding announcement u raised

how much did you raise

close to $1.4 billion uh which values the company now at $14 billion. [music] Uh, incredible, incredible. Um, yeah, let let's talk about where you're most uh excited about the application of robotics in the very very near term. Uh, there's so many different obviously angles, industries, kind of form factors, uh, that are exciting. But where where are you guys focused on today working with customers? Yeah, I think this is uh whenever in public folks talk about robotics uh on the consumer side, the excitement becomes okay, we have robot coming to our homes next month and I my robot will do X Y and Z. And people don't realize that robotics is not a new field like this is one of the oldest area in technology. It precedes computer and in fact it preeds AI

and robotics for 70 years. You see all these movies and even real world demos of robots but still we do not see robots around us and we begin to feel that robotics has not arrived yet. But if you go to any go to any big factory go to like car factories or Amazon you will see like tons of robots thousands of robots doing things manufacturing items etc. Now what's the gap? The gap is we cannot get robots beyond these beyond these demos or beyond these factory settings to consumer scenarios because the brain has been missed and that's what we bring to the table. Now in this setup it is a step-by-step approach. Even in factories even in enterprise the robots are there but they are in a red tape area like behind an area where you cannot go and if a human goes there they have to turn off the robot and that's where you start first you start to get these robots alongside people. So what we are doing right now even we are already deploying in variety of areas like uh pointto-point delivery systems even warehouses manufacturing but the change is we are doing it alongside you where people can come and interact and do their own work and the robot will go to the side in with randomized scenarios people is important because people create randomness and chaos and that's where AI needs to come in.

Yeah. How are you uh feeling about the supply chain? Are there any bottlenecks that you're worried about coming down the pipe? We've seen so many unexpected bottlenecks pop up in just the data center buildout. Talking about copper today, that was not on my bingo card. Uh what should we be worried about if we if we were going to be scaling robotics over the next couple years?

Yeah, I think this is exactly uh why we are scaling the way we are.

If you think about going like for every different application, there's a different hardware requirement. If you are doing something for data centers, you need a different kind of robot for doing things and specific other scenarios. And this is where getting all the supply chain all together at once, it takes a long time. But as I said earlier, robots are not a new thing. Industries have already been deploying them for variety of scenarios. The ROI has been less because you have to create a separate area, separate warehouse, customized scenarios for them. So we can piggyback on those supply. So what we the model we follow is we partner with hardware companies or even end users who have hardware and we provide our software as a service on top and that enables very fast go to market strategy for us and these robots which may be earlier in a in a secure area where humans cannot enter now can be used around in in outside alongside people and start to work.

Uh how much how much hardware do you guys actually build internally specifically for testing? I can imagine there's uh uh certain kind of like products where you're working just just on the software side working with a specific customer but then on the internal side for like you know having faster feedback loops I imagine you're uh building quite a bit as well

you can today what has changed the last five to seven years is you can purchase a lot of commercially available hardware uh and in fact every single robot hardware probably you have seen on the internet has made made its way to our office and we already have large number of them as long as the company sells them and gives us access to fully control the robot through our software. So we take the hardware we put the model is an end to model goes from direct from pixels in to actions out and it takes control of the whole hardware and makes it a skilled robot. Uh what what form factor uh do you think Rob uh what robotic form factor do you do you think will be the first to make consumers feel like robotics has really arrived? Right. You talked earlier about how it's like commonplace in a [clears throat] commercial setting, but it's happening like in a warehouse. No one's really seeing it and so they're not kind of like feeling the acceleration yet. But uh uh on the consumer side uh when do when do you think people will have that kind of like moment?

This is a very very very good question and I think it goes to the root of what people think of robotics when they say okay I I want a robot in my home. Okay let's say the most common thing in Silicon Valley like I want robot in my home in this year but when has it happened that you suddenly have have things in your home and you do not see them in your daily lives.

What about groceries? Like do you see robots in grocery store? What about Target? You go to you order something you waited for two hours for it to get picked up ready and then it's a pain to get even pick it up. What about hospitals? What about uh any of these uh consumer facilities which deal with consumers but are on the enterprise and that's where we believe you will begin to see robots first. It will become a commonplace thing and then you start to get them in your robot. Like if I give you a robot to you like today, you will probably use it and you will like it for for a few months. Then it will sit in your garage like a massage chair which I don't know [laughter] and then never get used again. So it has to first come become common in your daily life that you're not surprised. Okay, I have this in my home and and ready to go. And that's where we are enterprise is a stepping stone to the eventual consumer deployment. And yeah, I can go deeper into this. What is the other reason for this? But this is one of the practical reasons. The other reason is data

and that's can go for go into any.

Yeah. What what what what kind of uh bottlenecks are there on the data side that you guys are seeing?

Bottleneck as in [laughter] bottleneck when there is something in in high quantity and and it's coming through a squeezed hole. Uh but there is no data for robotics. This is the this is the the main uh the main issue. It is not one of those AI areas like as I said robotics was when the initial AI projects were done at MIT and other places they were done in the context of robotics. It has been 70 years chips have the gram has come down from a room size to like cannot even see that so tiny transistors robots is still not around them and the reason for that is unlike language or vision where you can build GPD like models LLM like models there's a data for robotics

and that's where the technical part becomes very critical so what we do today is we bootstrap someow you have to somehow bootstrap the process to deploy the robot in the on a site you cannot say I'll get a data deployment and uh but to deploy I need a problem like to deploy I need the brain to function for the brain to function I need data. So how do you break this problem? To bootstrap that what we do is we start with two such alternate sources. We learn by watching people. So learning by watching people act like uh there are tons of lifestyle videos on YouTube uh and Flickr and other sources. We watch them and have the robots learn from that. And the second source is simulation

and we put them together uh to transfer these results to the to the real world. And once they get deployed then we get we can establish this data privy where the more deployments increase the more data you get back and that increases your further your uh the performance of your model.

Very cool.

Makes a lot of sense.

Well, thank you so much for

Yeah, great to great to finally meet.

Yeah, great to finally meet. We're huge fans of Luke Metro. He's a great pet guy.

Yeah, we'd like to nominate him for employee of the month employee of the year.

We don't know anything about the other no insight, but he's the employee of the month in our

We think he's a great guy. He's a good [laughter]

of the month. Employee of the month every month.

Fantastic. Fantastic.

Uh well,

uh so great to finally meet you. Uh and congratulations on all the progress.

Have a great rest of your day. We'll talk to you soon. Cheers. Goodbye.

Public.com investing for those who take