Getting AI into perspective
Following on from my post a week or two ago, I wanted to continue the AI theme. Actually, I was going to write something on how I process the flows of information that I get and how that leads to new ideas for our Cities practice; but that will have to wait.
A couple of articles and a book I am reading reinforced the need to continue to plough the furrow of AI; its potential and limitations. In the last post I talked about the P vs. NP problem and the limitations of AI. This article by Monica Rogati (Machine Learning, AI and the Emperor's Vest) made me reflect further on hype vs. reality. Monica is keen to stress the need for balance in our expectations for AI. The reality is that we are constantly waiting to catch the updraft of the vortex created my mathematical advances in AI (deep learning algorithms, multi-agent planning, kaman filters, etc. ) and technical advances in the hardware and software environment. She rightly stresses the need for balance and humility and a more nuanced message about both the 'thrust' and the 'drag' of AI.
This duality is something that is a central theme of the fine book by Nick Bostrom on AI (Superintelligence). He talks about speed of take-off and recalcitrance at some length and the factors that will determine the acceleration, velocity and drag, that will ultimately determine the trajectory of AI developments. For anyone interested in understanding the potential for AI, I would really recommend taking the time to read this book. It is the most thoughtful and thorough piece that I have seen to date and I am sure it will become a classic in retrospect.
On the hardware side, a post on Engadget caught my eye about the new Fathom Neural Compute Stick that uses Movidius chips for plug-and-play neural compute power. This is already being used in the drone world to enable heavy duty AI for obstacle avoidance. On demand and mobile compute capacity is a great addition to the arsenal of AI practitioners and will enable all sorts of developments in the robotics world. I only mentioned the new NVIDIA Neural Chipsets a coupe of weeks ago. It seems that interplay between the hardware and software is already creating a vortex that will drive innovation. I just hope that the mathematicians keep pace and continue to push the boundaries of what we are capable of.
Whilst the innovation continues to power on, the focus will need to shift to us business practitioners, the regulators and a legislators. Hopefully, we will be able to keep up and continue to look for applications of the technology and convince our collaborators to adopt and embrace the new technologies. Likewise, it is important that regulators and legislators keep up to speed or get ahead of the curve to make sure that we do not hit a political and legislative brick wall.