Subscribe via Email

Friday, March 30, 2018

SIRLE 2018 - Stanford University

non-technical: suitable for any reader


Some previous work from Robot Magic was accepted into a symposium called SIRLE being hosted at Stanford University.
SIRLE (Spring Symposium on Integrating Representation, Reasoning, Learning, and Execution for Goal Directed Autonomy) focuses precisely on what the rather verbose title states. The goal is to create robots and A.I technologies that are fully embodied, that is they contain arguably all of the abilities that we as humans deem making us intelligent.
The symposium is really the textbook definition of what people picture when they think of "robotics research". I found it fascinating because of it's strong ties to some of the foundational AI conferences such as The Dartmouth Summer Research .
A lot of recent conferences have a very narrow focus, which is sensible in the academic context; but the aim, at least as I see, with SIRLE is to take a step back and look at things from a higher, and more fundamental level in terms of autonomy and autonomous agents.


Travelling to SIRLE



I knew it was warmer down south than in Winnipeg; but still, I was caught off guard by 26 Celcius at the airport. I removed my hoodie before I passed out, and honestly I would have removed everything else if it didn't mean getting potentially locked up by airport security. 

Here is a clip of the caltrain which travels across California. I was on my way from San Jose to Palo Alto.


I settled down in an AirBnB beside Stanford and was up until around 2am perfecting my slides. Since sleeping was apparently not in my agenda I spent another couple of hours running through the presentation and finally crashed by about 4:30am. I've given talks regarding competitions at IROS; however this is my first real talk at an academic event based on a published paper. I can't afford to mess this up.


SIRLE Conference day

Technically SIRLE took place between Monday and Wednesday; however I could only manage accommodation for a couple of days. Thankfully I received a student travel grant from SIRLE which helped cover some last minute expenses

I walked on down to the campus for about 7:30am.



The campus History Building 200
I was absolutely blown away at how beautiful the campus here is. I had some pre-exposure by working with Carnegie Mellon Last summer, which also has a beautiful campus. I felt like a house cat being let outside for the first time. Cat owners will understand: recalling memories of their cat clinging desperately to the lawn outdoors as if letting go will send them falling into space.
That was how I felt on this campus, metaphorically. Whether or not it was also physically is up to the readers imagination.

stanford campus is nice!
Another photo of the Stanford Campus


SIRLE is part of the AAAI Spring Symposium series which has 7 subgroups such as
AI and Society: Ethics, Safety and Trustworthiness in Intelligent Agents
Learning, Inference, and Control of Multi-Agent Systems
and of course, SIRLE, which is the category I am presenting in. All symposia were located in rooms within the Stanford History building 200.

Our group opened with a fascinating talk by Dr. David Aha, who is head of the adaptive systems laboratory with the NRL’s Navy Center for Applied Research. He described a model for goal reasoning agents that considers agents in which constantly deliberate about their goals over time. When an agent is always deliberating about its goals, these goals may change over time as new information is received in the environment. His model thus allows for adaptive goal generation in a wide variety of environments. Some examples included deploying on real world UUV (Unmanned Underwater vehicle), autonomous characters in Minecraft (yes, the game), coordination in air combat, and with disaster relief for heterogeneous teams.


This talk followed with a talk from Dr. Aha's colleague, Dr. Matthew Molineaux who outlined his formulation of a new agent-based planning model [found here]. He exclaimed that the existing RL model makes some unreasonable and semantically unpleasant assertions about the role of the environment (asking for what it wants) and the agent (learning directly from the environment).


Outline of the standard reinforcement learning model. An agent performs actions in an environment and receives an award, along with a state update that influences future actions. Image courtesy of analyticsvidhya blog.

He a proposed formulation -- which was open for discussion and improvement -- that extends the model by incorporating requests and explanations that allow agents to communicate amongst themselves and thus better coordinate.

My presentation was at 11:00am so despite really enjoying these morning talks, my mind was under constant tension. Everyone here set the bar pretty high with their contributions, so I was hoping my robot magic wouldn't come across as childish.
One thing I learned from the morning as well, was that you should customize your presentation to the conference. I spent a lot of time making presentation slides; but I hadn't realized until this morning what the focus of the series was on. Presenters before me were putting emphasis on goal-oriented reasoning and model representation; whereas my existing slides put an elaborate amount of detail into computer vision.


I on-the-fly adjusted my focus to give more emphasis to robot magic as a good medium for exploring high level problem representation and planning and as such I invested slightly less time than I originally planned, on outlining the specific implementation.

For those interested, here is a link to the symposium website which provides the full papers.

Also, here is a link to my full Robot Magic presentation.


Heading to San Francisco

I hopped on the caltrain and began departure to SF. The atmosphere was similar to Palo Alto; however I noticed a massive influx of tech people on the train. An ocean of MacBooks accompanied focused individuals with headphone-in-ear, avidly typing away at something. You also can't help but overhear at least one of the several simultaneous phone calls going on among people adjacent to you. I mean, I didn't actually take the time to record what people are saying; but it's important stuff nevertheless. Everyone has something important going on. I arrived in SF and the day flew by. I had some meetings to attend regarding an internship I'm starting soon (more on that eventually). I've been writing this blog post at different parts of the day from different locations. Right now I'm in Contraband, a coffee shop in south-of-market. The room has about 15 people. To the left of me are two girls and a guy designing a login page for their startup. The person to my right is solving interview prep questions on leetcode. This is exactly how I pictured SF to be.


QUICK DETOUR...
Ok so as I am typing this some song is playing in the background and the lyrics were:

"Then idle in some parking lot
Smoke half a smoke and ask
St. Boniface and St. Vital
Preserve me from my past"

I live directly next to St. Boniface, in Winnipeg, Manitoba, Canada. 
What on earth, are the odds that a song comes on while I'm in a random coffee shop in San Francisco that nearly references my address. I actually just stopped writing and went to ask the barista. He casually shrugged "Yeah, it's about Winnipeg".

So, for those from Winnipeg reading; be proud. We're on the map! The song is "Hymn of the Medical Oddity" by The Weakerthans.

You see, this blog comes in handy when funny things happen during my travels. Instead of messaging all my friends all the interesting endeavors that occur, I just dump all of the details here and those who are interested can read in their leisure. 

Anyway back to my regularly scheduled programming. Let's wrap things up as I'm meeting a friend at Pinterest HQ in an hour. 

Conclusion

This week was essentially a tour of California, starting in San Jose and migrating up to SF. I'm also noticing that on each of these trips I'm getting less and less time to actually write blog posts about them, and instead more time just having to keep up.
Some pretty major things coming up in the coming months. I'm excited and incredibly inspired by the people I've met this week. The culture here in SF embodies my mindset. People here are grinding hard, and everyone has a dream.

Quoting a gentleman from SF:
In Washington the dream is power
in L.A the dream is fame,
in San Francisco the dream is yours.

1 comment:

  1. Fantastic blog! Congrats on your sucess af the Stanford conference, and all the best in your future endeavours! It’s such a privilege and treat to be a part of your journey through these shared blogs! Next best thing to being there! Live the dream Kyle!

    ReplyDelete

Please feel free to give feedback/criticism, or other suggestions! (be gentle)

Subscribe to Updates via Email