PDA

View Full Version : Hi there, I'm new!



badshar
06-07-2014, 02:32 PM
Hello all,

My name is Saurabh and I am a college sophomore. I am studying computer science and might take up robotic engineering as a second major.

Since pretty much forever, I have been very interested in computers and technology in general. In fact, among my friends, I am usually the most technically literate and they usually ask me questions and help for technical thing, not trying to brag or anything, haha.

I remember a couple years ago, I started watching a show called, "Small Wonder", which used to come back in the 80's. The whole idea of humanoid robots amazed me. How you can program them and make them nearly human-like. When I watched Iron-Man, I was amazed yet again. The numerous types of robot, including one that had voice interaction and talked like a human. Since then, I have been doing a little bit research on robots, including being on my school's robotics team and competing in competitions.

I am excited to be a part of this community and hope to learn a lot of valuable things from this forum and the people.

tician
06-09-2014, 10:48 AM
I hate to rain on your parade, but the Iron Man movies are mostly CGI and technobabble, and a near complete denial of the laws of physics. No amount of rigid armor will keep the rapid accelerations routinely performed in the suits from crushing bones, rupturing blood vessels, stopping the heart, and eventually smashing the brain into a puddle of bloody goo within the skull. Just falling over from a standing position onto a concrete sidewalk can cause enough brain damage to kill a human. Fortunately for robotics researchers, most components in robots are designed to be significantly less susceptible to damage from mechanical shock and vibration, and are also far easier to repair/replace when damaged, than their organic counterparts.

Most current humanoid robots still have great difficulty walking over uneven terrain, and the demonstrations/interactions with humans only appear 'natural' under very limited circumstances that the programmers anticipated and were competent to implement. Go outside the design parameters and the systems break down, some more gracefully than others. The better systems have generalized recovery processes with some amount of adaptation/learning and/or asking humans for help (http://cs.brown.edu/courses/csci2951-k/papers/tellex14.pdf), while the worse try to do the impossible and create special checks and corrections for each and every edge condition. Natural Language Processing is one of the areas that will be essential to human-interactive general AI, but better handling of 'big data' and much more powerful, yet energy efficient, processing hardware will be at least as important. Watson (http://en.wikipedia.org/wiki/Watson_%28computer%29) required 90 servers (~80 TFLOPS and ~$3 million) for his run on Jeopardy, and is nowhere near general AI. The research projects of today are steps along the winding paths to general AI, but it is not here now and will not be here tomorrow. It will take a lot more time, funding, and creative humans to achieve general AI and useful domestic humanoid robots, so I really hope you do join some part of the process to make them happen.