I thought it would be valuable to take a step back, add some historical context to some important work we publicized recently at Bonsai, and to start to explore what
I thought it would be valuable to take a step back, add some historical context to some important work we publicized recently at Bonsai, and to start to explore what this could mean for the future of deep reinforcement learning (DRL). In contrast to some frequently cited achievements in DRL, these results show the tremendous business value that can be realized when this technology is applied to real world systems. To explore this achievement in greater detail I recently wrote a post over at insideBIGDATA titled “Deep Reinforcement Learning: From Board Games to the Boardroom”. In this article I lay out the trajectory DRL has been following so far, the obstacles it has encountered along the way, and explain where it may be going from here. Head over, take a read, and let me know what you think.
In this episode Andrew Vaziri speaks with Nicolas Economou, CEO of the eDiscovery company H5 and co-founder and chair of the Science, Law and Society Initiative at The Future Society,
In this episode Andrew Vaziri speaks with Nicolas Economou, CEO of the eDiscovery company H5 and co-founder and chair of the Science, Law and Society Initiative at The Future Society, a 501c3 think tank incubated at the Harvard Kennedy School of Government. Economou discusses how AI is applied in the legal system, as well as some of the key points from the recent “Global Governance of AI Roundtable”. The roundtable, hosted by the government of the UAE, brought together a diverse group of leaders from tech companies, governments, and academia to discuss the societal implications of AI.
Check out our application video for Battlebots Season 2, coming this summer on ABC! Looking Back I should tell you what the asterisk in the title is for. Our robot
Check out our application video for Battlebots Season 2, coming this summer on ABC!
Looking Back
I should tell you what the asterisk in the title is for. Our robot would have been the first BattleBot of its kind. The ABC casting committee loved the video and all our application documents were in order, but in the end ABC declined our application. While it is disappointing news, I would like to thank ABC for bringing BattleBots back, and for giving us some broad stroke feedback about why we were rejected.
So why did this robot get cut? First off, ABC needs to secure a variety of designs. On the application form there were about a dozen design archetypes. Right off that bat that means there will be some overlap for the 36 spots. Out of the 228 (wow!) applications that were received, apparently there were more full body spinners than expected, and we just got edged out. Secondly, they felt it was unlikely that an audience would care about the difference between remote control and self-driving. Perhaps taking the time to explain an autonomous robot is too dense, but, dang, I would have loved to introduce some ideas from robotics to the general public, or better yet go under the hood with Tested! In any case, perhaps the biggest problem with the application was they did not think it is possible to make a self-driving BattleBot that is any good.
I always knew technical feasibility would be a concern. The casting committee did not recommend including technical details in the application video (the video was supposed to focus on the team and the story). Even so, I wanted to pre-empt feasibility concerns by including footage to show the level of development in our CAD and software. ABC's robotics experts were not convinced, and unfortunately we had no ability to respond to their concerns until after they had made their decision.
Still, we had a phenominal team, and I don’t think you could make a better application for this idea short of having already made the robot.
So I will!
Looking Forward
This robot won’t be the first BattleBot of this kind. Nope, I will have to go with the generic untelevised store brand of “combat robot”. The goal is to have a robot ready for RoboGames.
Without the funding that BattleBots provides to the teams my ambitions will have to be a lot more modest. I will be scaling down the design to 60 or 120 lbs and I will be losing the complicated weapon mechanism. The focus will be on integrating the sensor and computer so that we can get everything tuned up to move autonomously as fast as possible. Furthermore, I will be using open source software and hardware whenever possible. I am excited to share the code, and designs. I will write posts and shoot videos detailing the fabrication of the robot, the writing of the software, and the fires testing. I see this project as a kind of no frills implementation, a jumping off point for anyone who might be interested in making an autonomous BattleBot of their own in the future.
If you would be interested in seeing this idea become a reality please leave your comments and questions below! I would also like to thank EandM and SICK who agreed to continue to sponsor this project by providing a state of the art laser distance scanner even though we wont be on TV.
Sign up on our mailing list to be notified when a new post is out. The next post about this project will include more details, and an FAQ drawn from your questions.
