We gave a lecture at Stanford University about how we are creating a novel AI agent.
One day in August, Stanford University made sudden contact and said “Could you give us a lecture on Virtual Human Agency, the AI project you at Couger are working on? ”
This AI project has been featured since 2018 at events in New York, San Francisco, Tronto, Berlin, Singapore, Hong Kong, Shanghai, and has also been covered in Japanese media.
Then, Stanford University offered us to come and talk about our project.
What is a Virtual Human Agent and what makes it interesting to all of the above, as well as Stanford University?
Simply put, it’s an “AI character”.
However, it’s not a normal AI. The following are some of its unique features:
・Senses of Sight and Hearing
・Recognition of Space and Objects
・Smooth conversation is possible
・Reaction in real-time
・Emotion and series of memories
Having a sense of “Sight” by object recognition, and “Hearing” by voice recognition makes it possible to react with objects or humans in the real world, as well as making interactive conversation.
Also, the agent can recognize objects at high speed and doesn’t need any wake-up word such as “Ok, Google” or “Hey, Siri” to start conversations.
Do you want to see what it’s really like? Then please watch videos below, and you should get it perfectly.
▶︎Video1 : Virtual Human Agent born from “Connectome”
・recognizing human with camera
・linkage with consumer electronics
・making conversation on the weather or calendar information
・move around between monitor and smartphone
・recognizing objects via a smartphone camera
・approaching humans when it finds them
・when we call the agent, it turns around and comes towards us
▶︎Video2 : “Life-size” Virtual Human Agent in action!
・moves to the room with the user
・when you turn off the smartphone, the agent appears on screen in real size
・gives us best solutions by linking with consumer electronics
▶︎Video3 : Virtual Human Agents on complex computer vision
・recognizes two people coming into the conference room
・recognizes that there are two bottle of drinks in the room
・suggests the nearest staff to bring two more bottles to the conference room
Having Emotions and Memories
A Virtual Human Agent has a special element that is described as a combination of Sight and Hearing (or more than that).
IT has “Emotion and Memories”.
The agent combines the memories of previous events and real changes of facial and emotional expression which effect communications.
Our goal is to create a more human-like agent.
Why we create this? Why is it so realistic?
The reason why we create Virtual Human Agent is that it would be a “Brand-new interface”
Our interface has been evolved steadily like above.
We discussed seriously what kind of interface should exist in a few years. The answer was “Human-like interface ＝ Virtual Human Agent”.
Nowadays, smart speakers can also make natural conversation but they don’t have a face (facial expression). People may feel it very inorganic and conversation without seeing facial expression is uncomfortable.
So, this is why we develop the agent to be “Reliable” and “readable”.
In addition, if we can’t see the face of the agent without its emotions, we can’t have a heart-to-heart conversation. Actually, this may give people a bad impression because we feel “We have no idea what it’s thinking about ＝ Strange”.
Virtual Human Agent, which is what Couger is developing, is not just a convenient machine to carry out people’s order. We want to create the agent to be more flexible to suggest the best solutions for the user by considering several situations.
The important point for users to accept the solutions naturally from the agent is “Reliability”.
Do the users trust in Virtual Human? Do Virtual Human act in a trustable way?
Emotional abundance and a good memory is the key concept for making a realistic interface.
※Couger do this project by collaborating knowledge form member who is good at Game AI, Data science(researcher from India), and also researching the structure of behavioural psychology, and emotions.
We have been repeating discussions with inquiring mind thinking about “Why do human react like this”, “How does brain work for judging the various cases?”
At last, the Virtual Human Agent will go to Stanford University!
On 11th October(Local time), Mr.Ishii(CEO at Couger) and Mr.Ishuguro(Blockchain Engineer) are going to make presentations on Virtual Human Agent at Stanford University.
Information about Stanford University
Website : Someone may think “Why does Blockchain Engineer give lecture on AI agent?”, but we would like to answer this question in next time.