Anant Agarwal on “Computers that Learn” – Keynote at TiEcon, 2012

Anant Agarwal, professor at MIT, Director of the Computer Science and AI Lab, and Founder of Tilera Corp. began his keynote address on Computers that Learn, at TiEcon, 2012 by stressing that computers suck up a lot of power and it is challenging to get power to them.  The current computing model is over 50 years old and is broken, said Agarwal.  The current model is procedural control model where computers keep running based on prescribed algorithms but do not learn along the way.  The problem will only get more magnified, as we get to possibly 1000 core processors by 2020.

How to fix this problem?  Agarwal shared his vision for Organic Computing that is based on lessons from biology.  The body regulates energy input and output based on the feedback loop.  For instance, when the body gets tired, it would work slower, run slower etc.  How does the body regulate the energy output to maximize efficiency?  Body gets feedback in various ways.  The new Organic Computing model can be built along the same lines to rethink computing, based on ODA or observe, decide, act loop.   First, such a system needs to be self-aware, where it can track the energy usage, in order to act on it.  Then it needs to be adaptive to respond to changing conditions, while maintaining performance towards prescribed goals.  This is not very far-fetched, considering that we already have machines that learn.  For instance, based on the pattern of the movies you request, Netflix will suggest other similar movies, because it learned about your taste and preference in movies, said Agarwal.  However, the challenge is that these computers will have to be intelligent learners in order for them to maintain performance, while learning and adapting to changing conditions, more like super athletes and less like common people.  New challenges of energy and complexity will require rethinking of 50 year old computing model and Organic Computing Model holds a lot of promise, said Agarwal.

, , , ,

  1. Leave a comment

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: