Robot Not a Dev? Pre-order Now
Not a Dev?

Misty Community Forum

Misty's personality development - current status

#1

I recently discovered that the group responsible for developing Misty’s personality has been let go at the end of 2018. This was the Brain Engineering group headed by Dan Grollman.
I really regret that this happened. To my knowledge Misty was the first general purpose robot project to have a goal of including a coherent personality. That is, the goal seemed to be for Misty to express her personality while engaged in carrying out tasks, exploring her environment, and so forth. This aspect of Misty seemed to be strongly supported by its founder Ian Bernstein.
At the last Community Forum one of the presenters responded to my question about the current status of Misty’s personality development by saying that it’s in the hands of developers, or words to that effect.
So I’m curious if there are developers with thoughts about how to go about this.
I’ll share my in my following post on this topic.

#2

So here are my thoughts about one path that might be taken to help Misty incorporate a coherent personality.

I think one way that Misty could most easily express a coherent personality that would differ among her cohort would be in Wander mode. The GitHub Wander sample skill could be a good place to start.

The easiest dichotomy to differentiate different Misty’s might be whether Misty was instrumentally or task oriented vs. person oriented. This correlates highly with gender differences but not exclusively so. That is, in interacting with a human is Misty focused more on finding out what she might usefully do? Or is she focused more on building a positive relationship with the human she is interacting with?

Back in September @Terrainpark had some interesting ideas about helping Misty develop some differential responses to the people she encounters. I think these two ideas could usefully be integrated.

#3

Hello there.

I am new to this group, but I run the Human/Robotic Interaction work at my company, Small Robot Company, and have worked on personality designs on AIs previous to this.

From our experience, robotic personality needs to be designed and tested before it can be coded. It is not as easy as ‘task oriented vs person oriented’, and I am extremely surprised that it these very basic differences are even associated with gender.

It needs to be made up of Misty’s habits, intentions and beliefs, which then filter down through environmental, organisational and location attributes to create the outcome phenomenons. These need to be mapped and planned before any coding is done, and be fundamental to the coding. A lot of work has been done, but us an others, to establish a user-centred process to achieve this. If it is possible for the Misty team to share the work that has been done on this, it would be excellent!

We have spent a long time on this with our agri robots because the quality of interaction has a huge impact on the takeup and adoption of our service.

I do hope that someone picks up on this for Misty. It seems a bit crazy that a social robot would not have these elements built in.

Ben

#4

I guess social norms vary across industries, but this public forum (that allows anonymous read-access) is likely not a good place to post something like this, much less with someone’s full name and role.

#5

Ive been wondering about a robotic personality for some time. I wonder how much of a basic personality could be sent out in an OTA update.

I understand Misty is a development platform. Im sure there must be many robot enthusiasts who would purchase Misty if it could do more put of the box. $2400 isnt an unreasonable price for a robot of this caliber.

As it stands, I dont know what my family and I would do with Misty in her current state. I read the ten year plan laid out for Misty Robotics. If my understanding is correct, Misty 3 wouldn’t be released until phase two starts after 3 years.

I was very excited to see new platforms by Nvidia and Google that could be plugged into Mistys backpack port.

#6

Normally I’d agree with you. In this case an entire group that was pretty central to Misty’s development was let go. The hoped-for replacement seems to be those of us who are developers.
If you search for “robot personality” and start reading from the first post in December, 2017 you can get some sense of the work that was going on in this area.
Also, I have yet to have anyone involved ask me to edit that post. If I do receive such a request I would be happy to comply.

1 Like
#7

@Ben1, I agree with your description of what needs to be done in developing Misty’s personality. If you search on “Robot Personality” you can see various thoughts on work being done at Misty Robotics.
In particular click on the tinyurl that is in that post. It describes the original “drive” model being used to develop Misty’s personality. This was later dropped because of the computation power that was needs, and replaced with a model of personality based on Eysenck’s work.
And I totally agree that sharing the work that has been done would be excellent. I am quite confident that there are hooks that exist that would be most useful for future work.

1 Like
#11

Hi Everyone! I’m delighted that this topic has taken off!

Let me address a few questions and comments in the hope to make our work on personality a bit clearer. First and foremost, work on Misty’s personality is still planned, and the groundwork that Dan and Michael laid has not been disregarded. As some of you may be familiar with, building robots is a complex process and we have been working hard to create one that is robust, sensor-rich, and easy to program. As manufacturing complexity has ramped up, we need to focus on getting the robot built and in your hands to help define what features and tools are needed to create your own personality. We have come a long way in validating what was a research project inside Misty Robotics, but the subtle distinction for this conversation is Misty will not have a fully fleshed out personality, but a programmable personality.

We have a path forward for more intricate features around personality and we’re still determining exactly which will be most important to deliver through software updates soon after Misty is released. To a large degree, the priority and direction of personality development will be determined by feedback and feature requests from those in the Misty Community.

To address @ben1 , We absolutely intend to test foundational personality aspects and have not yet come to a final decision on what those anchors are going to be (person, activity, task, etc), but we are hoping to do more in this arena after we ship!

In closing, I should point out one more important distinction: Misty is not a Social Robot but a personal robot built for programmability and rapid iteration. We sincerely hope that for some folks, that means building and experimenting with their own social robot, but having an answer for ‘this is what a robot personality should be’ is going to take some time for us, and you, to experiment with.

3 Likes
#12

Hi @Chris.
Thanks for your reply. I’m not sure whether this would be a features request but it would really help if we understood a bit of what tools or such are needed for Misty’s personality to be “developed”.
A brief response on why developers can’t be part of this now would also be helpful. I’m assuming that given the early work on Misty’s personality that there are hooks to doing this in Misty I as well as Misty II.

1 Like
#13

I really appreciate the additional responses to this query. Its great to know that this hasnt been sidelined. Im sure many people who become acquainted with Misty Robotics will be pleased to see this also. I cant recall any robots that had even a passable false personality.