Tuesday, December 1, 2015

Artificial Intelligence At Oxford

Ha! I'm (sorta) following in Izabella Kaminska's footsteps with a link from Oxford Today magazine planned for tomorrow.
From FT Alphaville:

Opening Pandora’s AI Box in Oxford
About three months ago, Dr Simon Stringer, a leading scientist in the field of artificial intelligence at the Oxford centre for theoretical neuroscience and AI, fell down some stairs and broke his leg. 
The convalescence period proved unexpectedly fruitful. 
Freed from the daily rigmarole of academic life, you see, Dr Stringer’s mind was able to wander. And so it was, when he least expected it, that the solution to one of the biggest challenges in artificial intelligence — the so-called binding problem — struck him out of the blue. 
“Of course!” he exclaimed in the style of Archimedes before him. “We need a top down approach, as well as a bottom up one!” 
Hitherto, Stringer’s team had differentiated itself from engineering AI approaches in the market by honing in on the idea that true artificial consciousness requires spatial and causal models to intelligently interact with the world around it. Simply put, an AI needs an internal imagination. Only then can it mimic the equivalent of a truly intuitive intelligence which relies on much slimmer real-time datasets. 
The presumption thus far, however, was that incoming data from AI sensors would need to be filtered through the neural network on a bottom-up basis — separate networks effectively battling each other out until only one (including its preferred action path) dominated. 
But, says Stringer, he now believes there’s a top down executive component to the brain’s functions, and this too must be accounted for if true consciousness is to be created. 
As he noted to FT Alphaville at a faculty dinner last week: 
For example, while the aforementioned engineering models rely on purely feedforward connectivity between neuronal layers, the cortex has extensive back-projections flowing in the opposite direction.Also real neurons in the brain communicate with each other by emitting electrical pulses called action potentials or ‘spikes’. 
These sorts of biological features, Stringer says, have been largely ignored by engineers working on AI. But he insists consciousness clearly arises from such a cortical architecture. Understanding this is what makes his Oxford lab different to the rest....

Before we get to tomorrow's post here's the current issue of Oxford Today with the most read feature being:

The humble philanthropist: How modest multimillionaire Lord Nuffield saved many an Oxford college