ROBOTS MIGHT NOT understand love, but they might be able to self-love like humans as OpenAI's Dactyl system looks to improve the dexterity of robot hands.
The company, which was founded by Elon Musk when he's not running his mouth on Twitter or sending Tesla cars into space, has created a system that can learn to be more dexterous through manipulating a cube in all manner of environments to effectively attempt to simulate how humans learn to use their hands.
OpenAI did this through using the ShadowDexterous Hand, a hand designed around human hands. This is added to its Dactyl artificial intelligence (AI) system, which uses the same general purpose machine learning algorithm as the AI system that learned to play Dota 2 like humans.
Through the use of cameras on the robot hand, the system tracks the hand's fingertips in real-time and tries to figure out the best way to manipulate the cube in any given situation.
The system learns using a machine learning training model called domain randomisation, whereby it carries out actions rather than repeated trial and error in a static environment; pretty much like how us simple humans learn.
"Dactyl learns to solve the object reorientation task entirely in simulation without any human input. After this training phase, the learned policy works on the real robot without any fine-tuning," explained OpenAI's boffins, noting that "dextrous manipulation using traditional robotics approaches" has previously been slow and "current techniques remain limited in their ability to manipulate objects in the real world".
But OpenAI has apparently got over this hurdle with its Dactyl and it noted that its work allows such grabby AI systems to be trained fully in simulations rather than require real-world training.
Such simulation enables Dactyl to figure out behaviours and hand movements that it wasn't strictly trained to do, again much like a lot of fleshy meatbags do when figuring out stuff.
There's no word on what OpenAI is going to do with the system, but we'd hazard a guess that such tech will lead to more human-like sex robots that can cop a feel in a natural fashion. µ
A surprisingly busy week in a quiet month
Measures just 15.75mm at its thickest point
Firm expects GPU sales to start drying up