Google buys Deepmind

[cosmicalstorm]

I wonder if this is the beginning of the AI-age or if it's just another round of AI hype.

Google has bought UK start-up DeepMind for a reported £400m, making the artificial intelligence firm its largest European acquisition so far.

DeepMind was founded by 37-year-old neuroscientist and former teenage chess prodigy Demis Hassabis, along with Shane Legg and Mustafa Suleyman.

The artificial intelligence company specialises in algorithms and machine learning for e-commerce and games.

Technology news website Re/code first reported the purchase price.

But Google declined to confirm the figure, while privately-held DeepMind was not immediately available for comment.

Major technology firms such as Google, Facebook, IBM and Yahoo have been increasingly focused on developing artificial intelligence as a new source of business.

Google for example, has been developing self-driving cars and robots, and in May announced a partnership with NASA in launching the Quantum Artificial Intelligence Lab.

The Quantum Artificial Intelligence Lab is aimed at using supercomputers and complex mathematical formulas to help improve aeronautical science and space exploration.

Earlier this year, the company bought military robot-maker Boston Dynamics for an unspecified sum.

The internet giant also hired futurist, inventor and entrepreneur Ray Kurzweil in 2012 to lead an engineering team focused on machine learning and language processing.

http://www.bbc.co.uk/news/technology-25908379

[Starglider]

The 'AI age' in the sense of IT automation becoming more and more powerful and robotics becoming viable for more and more applications has been underway since at least the late 90s (most commonly stated end of the 'AI winter'). In fact lots of generally available enterprise search, automated trading, marketing lead generation, product recommendation and even some game AI would be considered bleeding edge AI research in the 1990s. Self-driving cars and single-task robots sufficient to cost-effectively automate a lot of service/manual labor jobs are already enough to transform economies and make lots of money for investors.

General ('human-like') AI and seed (self-rewriting) AI are qualitatively different things that don't feature in VC pitches or even press releases as much these days. Progress is still being made on the bluesky stuff, but the hype seems to have refocused on near term applications of incremental development. Which is probably for the best.

[LaCroix]

Wait a second - first, they buy every company that is working on humanoid robots, and now, they start buying AI conmanies?

I feel my hackles rising, slightly...

[Lagmonster]

I recall reading somewhere recently that all of Google's recent acquisitions make sense if you consider their purpose, more or less, is to create Star Trek.

[Starglider]

their purpose, more or less, is to create Star Trek.

Create and monetise Star Trek.

[The Xeelee]

More like Weyland-Yutani. The owners have already invested in a space mining company.

[PREDATOR490]

How does buying AI related companies create Star Trek ?

As for monetizing it - What connection can be made between Star Trek and AI robotics ?

Granted, going with Star Trek has a high public profile but the opportunities are rather limited for practical usage. Data ? Voice Activated LCARS ?
If your going with wild speculative association then it sounds more like trying to create I,Robot or Skynet if your going for the dramatic.

Frankly, I find it unlikely this is going to end up being some grand scheme to create Skynet or Surrogates
I would be more inclined to see this as a push to make better data management systems and automated gadgets.

To which I say: Bring on the stuff that drive themselves but fuck off with the updated software that spams me with shit from advertisers based on everything I have ever bought, seen or done while it does so.

[Starglider]

How does buying AI related companies create Star Trek ? As for monetizing it - What connection can be made between Star Trek and AI robotics ?

I read that as 'the lounging around on the ship part of Star Trek TNG'. Which is voice-controlled everything, implausibly good voice-driven search & hypothesis/simulation, always on seamless audio/video communication, holodecks and replicators. Holodecks are an actual stated technical goal of AMD's GPU division; the major unsolved technical part is the force feedback*, everything else is refinement and cost reduction. Replicators are how the hype potrays 3D printing; this is about as excessive as the AI hype circa 30 years back, but as with AI real progress and more modest but tangible applications steam ahead.

* Obviously various kinds of force feedback peripheral exist and continue to be refined, but there isn't yet a practical concept for how to make a 'walk around and experience solid rocks, walls, chairs, ferns etc' system.

[cosmicalstorm]

Might this kind of attention from Google make AI seem more interesting to other players? Could this lead to an acceleration of AI research in general?
The cynical part of me thinks it will just cause players looking for easy investment money to slap 'AI' onto every product, as happened with Nanotech.

[Starglider]

The cynical part of me thinks it will just cause players looking for easy investment money to slap 'AI' onto every product, as happened with Nanotech.

Well, that's exactly what happened last time, but we got some neat software products out of it. Not to mention driving hardware R&D, see the Connection Machine, designed for AI research but was actually used to pioneer massively parallel supercomputing for engineering and bioscience applications. Of course we have approximately a million times more computing power now, massive repositories of training data (via the Internet) and lots of juicy new algos (including much more mature neuroscience for the biomorphic guys), which might make the difference...