Scientific Sense Podcast

Saturday, October 6, 2012

Reality bites

Recent news that the Chandra observatory has produced data showing the Milky Way is embedded in a vast halo of hot gas extending many hundreds of thousands of light years and possibly forming a bridge across the Local Group, may help reintroduce reality and less fanciful speculation in Astrophysics. The field has been dominated by those adept at concocting undefined constructs – dark matter, energy and flow - to account for the missing baryons and the inexplicable tilt of the universe for many decades. Such a simple explanation – that one has to look harder to find the thin veil – if proven correct, may inject a much needed reality check, with many positive effects.

Occam’s razor rules and those who forget it in any field are not progressing knowledge, just introducing noise to constrain it. One can never underestimate the power of simplification and this is especially true in a regime of exponentially increasing data, driving every scientist into analysis-paralysis and pattern finding. Empiricism and speculation without a mathematical framework are unlikely to be useful. The tendency has been either to take the status-quo equation as sacrosanct and reject observations that do not fit or to use observations to create hypotheses without a mathematical foundation.

Next time, perhaps, it is better to seek simpler explanations for exotic phenomena and badly behaving equations.

Tuesday, October 2, 2012

Green Brain

Recent news from the universities of Sheffield and Sussex (1) about a highly ambitious project to replicate a honey bee brain, in an attempt to advance the stagnant field of Artificial Intelligence, is encouraging. Brains, driven largely by instincts are likely more amenable to replication by currently available techniques. However, such replication is closer to automation than intelligence. Research in this direction is useful to build more intelligent automatons. Adding a layer of cognition to machines could be useful. If this can be considered different from the larger vision of AI – an ability to replicate the human brain in all its grandeur, we may be able to advance both fields faster.

The engineering concept of Artificial Intelligence has been stuck, attempting to connect brain replication with automation, for many decades. One of the primary reasons is that the structure and semantics of contemporary software are not amenable to modeling holistic phenomena. It is easier to build an airplane or a robot from component parts systematically. Engineers have been trying to extend this basic idea to brains with very little success.

Advancing contemporary AI techniques to creating brains with high programmability, such as the honey bee brain, is a useful exercise to advance robotics. But it is unlikely to advance our understanding of complex brains.

(1) 'Green Brain' project to create an autonomous flying robot with a honey bee brain. Published: Monday, October 1, 2012 - 11:07 in Mathematics & Economics