Google Buys UK AI Startup Deep Mind 113
TechCrunch reports that Google has acquired London-based artificial intelligence firm Deep Mind. TechCrunch notes that the purchase price, as reported by The Information, was somewhere north of $500 million, while a report at PC World puts the purchase price lower, at mere $400 million. Whatever the price, the acquisition means that Google has beaten out Facebook, which reportedly was also interested in Deep Mind. Exactly what the startup will bring to Google isn't clear, though it seems to fit well with the emphasis on AI that the company underscored with its hiring of futurist Ray Kurzweil:
"DeepMind's site currently only has a landing page, which says that it is 'a cutting edge artificial intelligence company' to build general-purpose learning algorithms for simulations, e-commerce, and games. As of December, the startup had about 75 employees, reports The Information. In 2012, Carnegie Mellon professor Larry Wasserman wrote that the 'startup is trying to build a system that thinks. This was the original dream of AI. As Shane [Legg] explained to me, there has been huge progress in both neuroscience and ML and their goal is to bring these things together. I thought it sounded crazy until he told me the list of famous billionaires who have invested in the company.'"
Re:No matter which "Deep Thought" ... (Score:5, Informative)
Legg (Score:5, Informative)
Shane Legg's research is pretty cool, since it deals with very sci-fi-like problems in a pretty rigorous way. For example, his PhD dissertation "Machine Superintelligence" approaches intelligence in a non-anthropocentric way, from the perspective of computability http://www.vetta.org/documents... [vetta.org]
More recently he's tried to define an IQ-like metric for comparing different AI projects and measure progress in the field http://www.vetta.org/2011/11/a... [vetta.org]
Re:Voice assistant (Score:5, Informative)
The kind of voice control Google is after (as in "the second-most intuitive interface") is hardly the same as the kind of voice control that is available today. The first would be able to interpret your intent as well as a human could, possibly better (filtering out noise, asking to clarify ambiguities rather than making assumptions). And it's nothing like the command line, which does no interpreting, refining or clarification at all; it just executes a limited set of commands exactly as entered, with no room for so much as a misplaced comma.
It's exactly like a commandline, which have been attempting to interpret their input for decades (most famously with http://en.wikipedia.org/wiki/D... [wikipedia.org] ).
The two reasons modern commandlines don't do this are 1) lack of effort and 2) that it's often a very bad thing. According to http://www.nhplace.com/kent/Pa... [nhplace.com] one of the motivating factors for defining Common LISP was to stop DARPA from rolling out INTERLISP, and therefore DWIM, across all their projects.
As for clarification, I run into this all the time when typing non-existant commands (thanks to the "command not found" program) or using undefined variables (thanks to GHC).