Meet the Humanoid robot 'EMO': The mirror-learning robot that can mimic human expressions shockingly well

Advancements in robot technology are one of the most successful achievements of humankind, but there’s a flip side to this coin, where continuous advancements also come as a threat to overpowering human beings and replacing them in their work.

Well, to be honest, the flip side is overpowering most people’s thoughts!

Recently, a video emerged where a robot with human-like facial features allegedly learned and mimicked just by watching itself in the mirror.

This again raises the unavoidable question, “Can we trust machines?”

EMO learnt to talk like humans by just ‘practising’ in the mirror

Engineers at Columbia University’s Creative Machines Lab created EMO, a silicone humanoid robot face powered by 26 internal motors for detailed movements. Instead of programming speech behaviours, they let EMO self-teach by facing a mirror, practising thousands of expressions, twitches, and corrections, like a kid discovering faces.

According to an article on Columbia Engineering’s official website, they then introduced the robot to the internet. EMO mapped audio patterns to its learned motions, linking sounds to lip shapes and timing. Hod Lipson, the lab director, wrote in a statement, “We had particular difficulties with hard sounds like ‘B’ and with sounds involving lip puckering, such as ‘W’.” He added that more practice would boost precision.

Lip-sync humanoid robot was put to the test

Researchers tested how real EMO looked by showing its videos to over 1,300 people. They compared the mirror-learning method against simpler ones based on sound volume or face markers.

Shockingly, EMO came out way ahead, matching near-perfect lip sync.

What is the way forward?

EMO opens doors for robots in caregiving, teaching, or companionship, where lifelike faces might make interactions feel natural and comforting.

Although the robot managed to master convincing appearances without grasping language, emotion, or intent, just by self-observation and copying online videos, this is like copying humanity superficially, without true comprehension.

Faces are supposed to signal trust, but this suggests otherwise.

However, realistic but emotionless faces may completely undermine trust and could lead to deepfake fraud or allow manipulation where caution is necessary, making it harder to distinguish genuine human interaction from robotic imitation.