I like my AI like I like my air: invisible and useful
If you're my age (or your parents are just old souls), you've probably heard the old adage: children should be seen and not heard.
This is obviously an outdated and barbaric concept. Unless you apply it to AI.
Useful robots not withstanding, I'm here today to argue that the best AI, the kind that's going to propel us into the future, should be seen and not heard.
Or, rather, it should be neither seen nor heard.
AI's like electricity that way. I want to know it's there, I want to know it works, but I almost never want to have to think about it.
Most of what's useful in the machine learning world is what I call "backbone technologies." That is, we all know that Microsoft Azure is a big important cloud-based AI system. It's arguably Microsoft's most important product that isn't named Windows.
But Azure hasn't been anthropomorphized or characterized. It's just there, it does stuff, and that's cool.
And when it comes to robots, I think it's important now and in the future to ensure the general public views them solely as machines.
It's says something warm and kind about our species when we learn that people got upset when they saw the Boston Dynamics technicians appear to be shoving and kicking their robots.
Aww, aren't we great? We have the capacity to care about anything.
But what happens when robots permeate our society and our human penchant for emotion draws us to do incredibly stupid things on their behalf?
If you saw a toaster fall onto a subway track, you wouldn't risk your life to save it.
But a certain percentage of the population is completely duped by startup hype, big tech lies, and media hyperbole.
The first time someone dies trying to save a robot, the design paradigm calling for humanesque machines becomes a fatal one.
We don't need machines to become more human-like. We just need them to be useful.
Tales from the arXiv archives
Adversary-Agent Reinforcement Learning for Pursuit–Evasion
We got a runner!
The art of pursuit and evasion is not something you can pick up overnight. There's game theory, tactics, the fog of war, and myriad other unfolding considerations in a real-time strategy environment.
And I'm just talking about playing StarCraft II.
I came across this research paper on the 'Xiv (that's what us cool folks call arXiv) and it was a fascinating read.
Basically, researcher Xun Huang came up with a way to train AI to evade capture using a mini-game built into a StarCraft II AI training environment.
Yes, Neural covers AI and quantum computing — but we're so much more! We've got one eye on big tech, one on the startup scene, and a third robotic eye focused on politics and government policy.
And that's exactly why we're proud to introduce Neural at TNW Conference, featuring an amazing speaker list with people who work and create in a variety of domains.
From Unity'sDanny Lange– a legend in the gaming industry – to Shopify'sElla Hilal, we're hosting some of the AI/ML community's most important thinkers. And, as a technology journalist, I can't wait to see and hear whatElizabeth Bramson-Boudreau, the CEO and publisher of MIT's Technology Review, has to say.