A Really Fabulous Lecture We Saw Today on the Topic of AI, Computer Vision, Language and Robotics: Deep Learning.
My wife and I went to the Princeton Plasma Physics Laboratory's Ronald E. Hatcher Science On Saturday Lecture Series this morning.
The speaker and her topic was Dr. Kristin Dana, giving a lecture entitled: Training Robots: Deep Learning for Embodied Artificial Intelligence.
Before attending the lecture I thought this one would be the least appealing to me, but, as Dr. Dana is a wonderful speaker and an incredibly aware scientist as well as clearly a great teacher, I think it is one of the best lectures in this series I've seen over many, many years. It will be on line in a few weeks in the archives, and I recommend it.
AI is a very controversial topic here, on which I may be swimming against the DU stream, as I have done for many years with my views on nuclear power, although in that case, the current against me with respect to nuclear issues is clearly slowing, a good thing in the face of a collapsing atmosphere, maybe too little too late and the failure of the antinuke's so called "renewable energy," but well...
Dr. Dana is a very aware thinker, and the audience raised many questions touching on the ethical and sociological issues with AI and robotics. She certainly didn't come across as all "peaches and cream" about the topic. During the Q&A many people raised points about the ethical issues. (I wish I had asked the question I asked differently.)
Others raised the environmental issues, which are very real. The silver lining, for me at least, is that it is driving the development of nuclear power, something I regard as essential to saving the planet. (Nuclear issues were not discussed in this talk.)
I am not opposed to servers and computational power, while recognizing that of course these things are subject to abuse. (I am mildly amused when I read complaints on DU, a server dependent website.)
Like all human technologies it has both positive applications and others which are dangerous, even to the point of frightening. One of the interesting points was that people developing AI often don't understand exactly how it works on a deep level; they are often surprised that its capabilities exceed their expectations.
The question of course is kind of 1950's science "fictioney" to coin a phrase, the question being, will humans make themselves obsolete? After the lecture, I kind of understand the risk on a deeper level.
Again, the talk will be on line in a few weeks. I recommend it highly. It raises questions and makes one think.
I trust you're having a pleasant weekend.
= new reply since forum marked as read
