fbpx

Tag Archive for robots

Neural nets and modular capabilities

April 19, 2015 future, Writing Comments Off on Neural nets and modular capabilities

http://journals.plos.org/ploscompbiol/article?id=10.1371%2Fjournal.pcbi.1004128

This abstract, as it was originally presented via one of the science aggregation websites, was about robots.

Except it wasn’t.  Not really.  I mean, “robots” is the end goal that most people will understand, but what these people are examining is the way that memories are written and rewritten.  They are examining adaptation of the learning process and how that can be applied to the idea of building neural networks.

Right now, when we build an AI, whether it be to assemble cars or act as an enemy in a videogame, it’s very task based.  In an ultra-simplified form you get, “If this happens, you deliver that result.”

But this not only leaves holes in the logic (because what if THAT happens and you didn’t think of it beforehand?) but it makes the programming rigid.  The AI can only operate according to the rules you have set, so if a dog runs into the automotive assembly factory, or a player decides to sneak around the left side of the building instead of the right, you end up with a broken situation.

There is a risk (seemingly) that as we continue to develop neural network based AI’s that this will get coded in there too.  Not always on purpose, but because that is familiar ground.  It is something we can test easily.  It is something we can codify and deliver a clear result that can be shown to colleagues/investors, etc. to keep the funding and interest going.

But in order to truly make a neural net efficient at learning and executing new tasks, it’s got to retain the old tasks.  It’s got to use the knowledge it already has the learn the new stuff even faster (rather than having it handed to them in the form of a programming block).

 

Tiny Minds

February 5, 2015 future, Writing Comments Off on Tiny Minds

 

In one of Isaac Asimov’s earlier robot books, he takes a moment to explain why you might have a bipedal robot, shaped like a human, looking (roughly) like a human.  One of the reasons suggested was that they could be generalists, much like humans are.  They could be reprogrammed to drive a tractor, or climb a ladder, or ride a bike, thus putting all the cost in to the programming (software) rather than into developing custom hardware for every application.

AI’s are likely to follow a similar path.  In game development we already have limited “AI’s”, pieces of software that serve as bad guys, that make alterations to the game based on interactions with the player.  In fact, most games have to cripple these, so as not to make the overall game too challenging for players.  They “think” insofar as they receive input, check it against a set of parameters (and sometimes experiences) then change their behavior to suit. Kind of like the cat getting sprayed with water when it claws the couch.

When most people think about AI they think about something along the lines of Azimov’s positronic brains, a human analog, constructed in mimicry of human thought processes and morality.  The real truth is that AI’s will likely be specialized through an evolutionary process much like the one described in this article.  They will be really really good at doing one (or a small suite) of relative tasks.  They will be widely varied (one AI who robocalls you to find out your politics, one AI who cuts your grass, one AI who drives your car).

The REALLY clever companies out there (I’m eyeballing GOOGLE for this) are initially going to follow a similar pattern to what you see now with their suites of Apps.  You will find yourself allying with one brand or another when it comes to your AI’s because they will all be able to share data smoothly, your lawn care AI can talk to your shopping AI for more fertilizer and your shopping AI can talk to your nutritionist AI to see if you need a fresh shipment of Soylent ordered in).

 

 

« Older Entries   Recent Entries »