Google doesn't need anymore money, thank you very much. It's fine that they 'waste' it on research. Much like ol Elon.
Nonetheless, I think they need to think about doing something with less potential for serious problems. I found the phrase
We never told it during training, ‘This is a cat,’” Dean told the New York Times. “It basically invented the concept of a cat.”
To be the scariest thing I've read all day. It did that by parsing YouTube. That was the first attempt to parse YouTube with 'Deep Learning".
I do not want to be around when it finally figures out about 4Chan.
My OMG moment came when I read
Nobody is saying that this system has exceeded the human ability to classify photos; indeed, if a human hired to write captions performed at the level of this neural net, the newbie wouldn’t last until lunchtime. But it did shockingly, shockingly well for a machine. Some of the dead-on hits included “a group of young people playing a game of frisbee,” “a person riding a motorcycle on a dirt road,” and “a herd of elephants walking across a dry grass field.”
because looking at those images made me realize the machine basically trained itself to do couple two domains of knowledge that even experts in language acquisition and image recognition only partially understand.
That's just flat out amazing.
The other part that got me going "Wow" reads
The neural-net system was left to its own deep learning devices to learn game rules—the system simply tried its hand at millions of sessions of Pong, Space Invaders, Beam Rider and other classics, and taught itself to do equal or surpass an accomplished adolescent. (Take notice, Twitch!) Even more intriguing, some of its more successful strategies were ones that no humans had ever envisioned.
As an old-timer (older than Dean which makes me feel like I missed the boat by spending so much time earning a doctorate in the humanities), I wanted to know precisely what successful alternative strategies DeepMind had devised in which games.
I mean, besides being completely fucking cool, that shit is like gothic scary.
The end of the article where Hassabis notes that humans should never spend any time wondering what book they should read next made me think of Richard Powers’ Galatea 2.2 which is an incredible read about the attempt to build an AI capable of passing a Master's exam in English Literature. Not as nerdy as the /. might like but it raises many of the important questions that we face as machines increasingly become able to make autonomous decisions based on (as the article calls it) "unstructured data".
I'm really glad to hear DeepMind has formed an external board to monitor the progress of its development and while the composition of that board is secret, I do think the product of its deliberations should be made public. In any case, it won't be too long before the US government (or the government of whatever country DeepMind cares to be in) will consider it an issue of national security and categorize AI and neural net technology as a munition or whatever it takes to get greater insight into what DeepMind and companies like it are actually building.
