Have we cracked AI?

This page was updated July 2023 to reflect recent advances in our AI platform.

We do robots. True-to-life human characters that make jaws hit the floor and cause delight to ripple through an unsuspecting crowd. 

You might think we’ve created the worlds greatest artificial intelligence. However robots aren’t only AI and AI isn’t only about robots.

AI is usually just computer code, which isn’t as dramatic as our lifelike characters. But it’s used for speech recognition and facial/object recognition, or to run large language models like GPT 3 which our robots can be programmed with.

Artificial General Intelligence, AGI – the kind depicted in films like Her, Bladerunner and 2001: A Space Odyssey – sadly does not quite exist yet.

Your robots feel so real?

If you’ve spoken to one of our robots, you might well have enjoyed a great conversation. But you may of been speaking to a human through our innovative telepresence software, TinMan. Human intelligence will still beat artificial intelligence especially where visual cues are important.

All our robots run on the same Tritium software framework, which we can customise with new programs so that every robot-human interaction feels unique. We currently support several large language models including GPT 3, 3.5 , 4 and Chat GPT.

Speech recognition

Can our robots hear? We now include automated speech recognition (ASR) on our robots as standard, it converts sound to strings of text. It’s still difficult to interpret sarcasm, tone of voice or context, but ASR can nevertheless help create incredible interactions between people and our robots. 

Our current favourite software solution for ASR is Whisper, it supports many languages including all common European and Asian ones.

Computer vision

All versions of Ameca feature HD eyeball cameras, this feature is available for Mesmer robots too.  Camera images are fed to computer vision, (CV), software to find faces and objects.

Our default configuration, uses the Google MediaPipe framework for face detection, and robots will try to make eye contact and track people out of the box.

If you have a particular CV program you want to use, we can get it running on our Tritium framework.

Conversation

Chat solutions have improved vastly with recent advances in large language models like GPT 3. They do very well recalling large amounts of factual knowledge and can be trained with a custom ‘personality’.

However interacting with a real human through a robot using our telepresence software, TinMan, can still have advantages, like fewer mistakes and better personality.

We can provide pre installed solutions for automated chat interaction, and customisation of AI personalities. 

Voice quality

A raised eyebrow from our robots can be worth a thousand words. But when it comes to how our robots sound, we can give you three options:

  1. Recorded human voices which are truly lifelike, but inflexible for dynamic interactions.
  1. Computer synthesised voices which work in many languages and generate voice ‘on-the-fly’.
  1. Both human and computer synthesised voices, mixed as needed

Looking for anything else?

Tell us about your interaction ideas and we will look to bring them to life. Or if you have one of our robots and want to try your own ideas, the API is open for you to experiment.