We just finished assembling the hardware, specifically: - 13 Dynamixel servo motos, of which 6 dual motors, with 19 combined DOFs - 36 3D printed PolyCarbonate parts - Jetson TX 2 4GB - Open CR 1.0
We now want to develop some basic apps for Teddy. What would you build on top of it?
> Teddy’s Intelligence comes from a GPT conversational model and its custom-developed skills.
It's not going to be the sort of intelligence people are expecting, GPT is still just an algorithm mimicking understanding. You need to manage expectations better.
> Teddy’s freedom of movement comes from its 13 reactive Servo Motors with 19 DOFs.
In the picture it shows the robot holding a bottle. It won't do this very easily at all, especially a smooth bottle.
> Teddy’s Empathy comes from mood-tracking the owner and trying to influence his mood through AI.
Handy-wavy use of empathy - this has not yet been solved.
> We now want to develop some basic apps for Teddy. What would you build on top of it?
Unfortunately you will never get this thing to walk on all surfaces reliably. Bipedal walking is exceptionally difficult and your robot is going to fall over - a lot. Those Dynamixel motors will give out really quickly (the large gears). Your customers are not going to continue paying $100 a month after the first month when this thing breaks.
And this is the least of your worries. There are problems I am sure you are yet to even consider. For example, this thing weighs 2.3kg (5lbs). The motors need to at least be 15kg/cm (I'm guessing AX12). Their stall current will be upwards of 1A. 10 servos in stall condition (legs trapped for example) could be drawing 10A. But you can switch off the motors under special conditions? Then you need to handle back EMF (which will fry all of your motors). This thing is potentially a fire hazard.
My recommendation is to pivot towards a teddy puppy. Walking is easier (larger numbers of ground contacts), less motors (cheaper), less expectation of intelligence, etc. You can save on production costs and ship with a cheaper SBC, smaller battery (dog walk motions can be more efficient), reduce the cost of your platform and get more of your monthly instalments before the thing requires maintenance.
I'm sure you couldn't offer the robot for $100, but does everybody you're trying to interest understand that.
I was also commenting on the "first humanoid" statement, so with these issues, I'd suggest you spend a bit of time working on your language and marketing and making sure you're being clear. Don't let little issues like this destroy your credibility and harm your product.
You can be clear AND get interest.
I hope my comments aren't coming across as negative, that's not the point, the point is to try to help because these are some issues that I find are destroying your credibility with me.
Have you looked / interacted with Pepper and Nao robots? In my experience, Aldebaran Robotics tried hard to go down the "lifelike robot" route, and the issues they encountered are real and hard problems.
A big problem with body control was lack of touch / impact feedback. Do you have plans to integrate sensors on the robot's 'skin'?
Modern ML based perception is expensive (audio, visual, NLP, etc). Modern ML-based full-body planning even more so. Do you intend to go down a hand-crafted route for walking, gestures, grasping, or do you intend to do it all with ML?
Btw, like others have said, it's a bit creepy looking. I didn't realize someone could make a teddy bear and somehow make it look cold and murderous but bravo. If you're going to make a teddy bear, at least make it look soft and cuddly. And give the eyes some life to them please
App ideas:
Diagnostics and Repair: to it can help you take itself apart and put itself back together.
BearSwarm/BeoBaer: connect multiple Bear together into a cooperative unit
Bear8s: load balance those units!
WarBear: If these work and only cost $3000 a pop, you can expect the Army to want some.
Fetching important medications, water and nourishment for those with limited mobility; Answering door; going outside to check on noises; other telepresence applications with VR linkup to the humanoid's cameras and other sensors.
Which GPT are you using with Teddy? When do you expect you will have the first live demo videos? What do you use for conversational technology - STT/TTS, NLU?
Also, if you could please share your social media so that I can follow your progress. Thank you.
> What would you build on top of it?
I'd go for something big and heavy...
It would actually be interesting to know how to interpret the usage of AI in the context of GDPR. Since the purpose of the processing must be pre-determined, yet with AI the data collected may be used for a purpose determined by the AI. Also "mood-tracking" will certainly fall under article 9 - special categories. You might end up in a situation where you have to delete all previous data and renew the consent every time someone installs a new "app".
> How does Teddy respect my privacy? > You can tell Teddy to go to sleep, and in that mode the camera is off. Teddy will only wake up when you ask him to.
Oh good. So it is always listening. That's great. At least the form advertises the function.