AI robots could outgrow smartphones, says UniX AI CEO

Ajouté : 2026-04-30

[00:00:00]Uh, so our company, Unix AI, we uh we were founded in April 2024. We are a embodied AI company that are focusing on robotics manipulation, and our main downstream application is in the service scenarios, including we call it cleaning, security, retail, and this kind of stuff.

[00:00:21]I think since the last the the Q4 for 2025, our robots are was entering into the mass production period. Uh, so roughly speaking, every every month we are delivering 100 to 150 robots in the real real application scenarios in the field.

[00:00:39]So first I would like to talk about Unitouch that you developed. It's a tactile foundation model. So put simply, what does it do exactly? So basically, in the previous age of robotic manipulation, people only use cameras, 2D cameras or 3D cameras for capturing capturing some videos for for guidance to guide where the robotic arm should be. You cannot tell how much force you need to press or pick something using only the vision signal.

[00:01:11]Uh, so tactile signal is basically we we mimic how the human are are pressing the object and and the objects are giving a counter feedback to the human finger itself.

[00:01:23]The actually the function of the Unitouch is that we use a vision as an intermediate modality that actually connects tactile to vision, language, audio, and and other modalities that may not be naturally paired with tactile signals.

[00:01:39]To be concise, after we have this model, we are building the connection between tactile signal and vision signal, and thus we can generate tactile signal even without touching the object. And what is your design philosophy? Can I say that your design philosophy is utility or practicality over humanoids? I would say yes.

[00:01:59]So the reason we have this foundation model is that you will found that actually tactile sensor is a very immature sensor compared to the vision compared to the camera or other sensors.

[00:02:13]It's pretty pretty expensive.

[00:02:15]So when I was a PhD at Yale just 2 years before, the cost for one tactile sensor is roughly $1,000.

[00:02:23]And and there's a gel on the top of the sensor. Normally that can only last for 3 to 5 years. And that makes the tactile signal and tactile sensor really hard to commercialize.

[00:02:34]Uh, so our my design philosophy for this model is that can we have can we use the tactile signal without this very expensive uh but very elastic inelastic tactile tactile sensor. So this is the philosophy behind the model. Humanoid robot is a very wide conceptual conceptual definition. Uh, it's not only the robot that have two legs and two arms that we call a humanoid robot. Uh, but also the robot that actually works like a human. We call it embodied intelligence a humanoid robot. And how big is the market for embodied artificial intelligence you mentioned just now?

[00:03:19]Uh, I would say in theory it is a it is a market that is basically a new industrial revolution that basically in in in the eventual setting it can basically do everything. Cuz it is the ultimate uh it is the ultimate version that AI is used in the physical world.

[00:03:41]So in theory it is a it is very it's very large one. It's a countless market.

[00:03:46]But I think for now we are only focusing on a very very small subset of the market. But even though even though that is very huge market. So take one of the job we are doing for example, we are we have a robot that can act as a uh safeguard in the in schools, buildings, hospitals, and and and home and houses.

[00:04:09]Merely in China we have I think 8 to 10 millions of people doing this kind of job. This is only one of the very small pieces of the of the task that a humanoid robot can do. So in general I would say embodied intelligence market is a is a countless market.

[00:04:26]Well, Fred, I think many people are wondering just how close are we to robots that can really understand and interact with the physical world like humans do?

[00:04:37]Do we have a timeline?

[00:04:40]Uh, so we define the autonomy or the the autonomy of embodied intelligence into actually couple levels. I would say the the ultimate level we we we call it internally a L5 level. It's basically a robot can act as a Superman or or or Jarvis. I don't know whether you've you've seen the the movie the Marvel movie Iron Man.

[00:05:04]This is the ultimate goal. Uh, so in this ultimate scenario the the actually the the robot is capable of understanding natural language of humans are saying and they can do exactly what what what human is asking you to do in various couple of environments. No matter you have the generalization in terms of skills, uh in terms of environments, and in terms of the in in terms of the objects. And this is the ultimate goal. I would say we still need a lot of years. But we are if we are pushing the autonomy back a little bit, we call it L4 autonomy, where I think we don't actually push the robot to do everything that human can do. But in this kind of autonomy that we will say we will say that human can the robot can do some can can do some task in some limited scope. But in a relatively generalized generalized setting. So for example, we are acting as a night guard but for different buildings, for different for different scenarios, for different lightning conditions. For for example, you want open the door for actually different kinds of doors we can open. I would say for this kind of scenario we are actually approaching this point. I would say in in a year or half or something like that. And I would say that if the robot can entering the L4 level of autonomy, it has huge commercialization value.

[00:06:41]Maybe maybe you don't have a robot that cannot kick that can do everything. But you will have couple robots that are specializing in scenarios that are experts similar to human that that it will create a very very large commercialization value.

[00:06:56]Well, Fred, are you betting on a future that robots will be like smartphones that everyone will have one in their households?

[00:07:06]Uh, so so actually I even have a bet that is larger than a smartphone. So for smartphone is that everyone has a robot.

[00:07:15]Only is it to see is it to customer side as a consumer business. But for a robot, if we are going to the ultimate phase, it's not only a a stuff or it's only not only item that every people should have in their house. It also enters into industry. It enters into the service field. It also enters into every field that actually a human is working right now. So I would say I would take a even even higher bet than a actually a car or a smartphone in the ultimate goal.